root@jetson:/jetson-inference# cd jetson-inference/build/aarch64/bin root@jetson:/jetson-inference/build/aarch64/bin# ./imagenet.py images/orange_0.jpg images/test/output_0.jpg jetson.inference -- imageNet loading network using argv command line params imageNet -- loading classification network model from: -- prototxt networks/googlenet.prototxt -- model networks/bvlc_googlenet.caffemodel -- class_labels networks/ilsvrc12_synset_words.txt -- input_blob 'data' -- output_blob 'prob' -- batch_size 1 [TRT] TensorRT version 8.0.1 [TRT] loading NVIDIA plugins... [TRT] Registered plugin creator - ::GridAnchor_TRT version 1 [TRT] Registered plugin creator - ::GridAnchorRect_TRT version 1 [TRT] Registered plugin creator - ::NMS_TRT version 1 [TRT] Registered plugin creator - ::Reorg_TRT version 1 [TRT] Registered plugin creator - ::Region_TRT version 1 [TRT] Registered plugin creator - ::Clip_TRT version 1 [TRT] Registered plugin creator - ::LReLU_TRT version 1 [TRT] Registered plugin creator - ::PriorBox_TRT version 1 [TRT] Registered plugin creator - ::Normalize_TRT version 1 [TRT] Registered plugin creator - ::ScatterND version 1 [TRT] Registered plugin creator - ::RPROI_TRT version 1 [TRT] Registered plugin creator - ::BatchedNMS_TRT version 1 [TRT] Registered plugin creator - ::BatchedNMSDynamic_TRT version 1 [TRT] Could not register plugin creator - ::FlattenConcat_TRT version 1 [TRT] Registered plugin creator - ::CropAndResize version 1 [TRT] Registered plugin creator - ::DetectionLayer_TRT version 1 [TRT] Registered plugin creator - ::EfficientNMS_ONNX_TRT version 1 [TRT] Registered plugin creator - ::EfficientNMS_TRT version 1 [TRT] Registered plugin creator - ::Proposal version 1 [TRT] Registered plugin creator - ::ProposalLayer_TRT version 1 [TRT] Registered plugin creator - ::PyramidROIAlign_TRT version 1 [TRT] Registered plugin creator - ::ResizeNearest_TRT version 1 [TRT] Registered plugin creator - ::Split version 1 [TRT] Registered plugin creator - ::SpecialSlice_TRT version 1 [TRT] Registered plugin creator - ::InstanceNormalization_TRT version 1 [TRT] detected model format - caffe (extension '.caffemodel') [TRT] desired precision specified for GPU: FASTEST [TRT] requested fasted precision for device GPU without providing valid calibrator, disabling INT8 [TRT] [MemUsageChange] Init CUDA: CPU +203, GPU +0, now: CPU 226, GPU 3456 (MiB) [TRT] native precisions detected for GPU: FP32, FP16 [TRT] selecting fastest native precision for GPU: FP16 [TRT] attempting to open engine cache file networks/bvlc_googlenet.caffemodel.1.1.8001.GPU.FP16.engine [TRT] cache file not found, profiling network model on device GPU [TRT] [MemUsageChange] Init CUDA: CPU +0, GPU +0, now: CPU 226, GPU 3456 (MiB) [TRT] device GPU, loading networks/googlenet.prototxt networks/bvlc_googlenet.caffemodel [TRT] device GPU, configuring network builder [TRT] device GPU, building FP16: ON [TRT] device GPU, building INT8: ON [TRT] device GPU, workspace size: 33554432 [TRT] device GPU, building CUDA engine (this may take a few minutes the first time a network is loaded) [TRT] [MemUsageSnapshot] Builder begin: CPU 292 MiB, GPU 3387 MiB [TRT] Applying generic optimizations to the graph for inference. [TRT] Original: 141 layers [TRT] After dead-layer removal: 141 layers [TRT] Convert layer type of loss3/classifier from FULLY_CONNECTED to CONVOLUTION [TRT] Removing shuffle_between_pool5/7x7_s1_and_loss3/classifier [TRT] After scale fusion: 141 layers [TRT] ConvReluFusion: Fusing conv1/7x7_s2 with conv1/relu_7x7 [TRT] ConvReluFusion: Fusing conv2/3x3_reduce with conv2/relu_3x3_reduce [TRT] ConvReluFusion: Fusing conv2/3x3 with conv2/relu_3x3 [TRT] ConvReluFusion: Fusing inception_3a/1x1 with inception_3a/relu_1x1 [TRT] ConvReluFusion: Fusing inception_3a/3x3_reduce with inception_3a/relu_3x3_reduce [TRT] ConvReluFusion: Fusing inception_3a/3x3 with inception_3a/relu_3x3 [TRT] ConvReluFusion: Fusing inception_3a/5x5_reduce with inception_3a/relu_5x5_reduce [TRT] ConvReluFusion: Fusing inception_3a/5x5 with inception_3a/relu_5x5 [TRT] ConvReluFusion: Fusing inception_3a/pool_proj with inception_3a/relu_pool_proj [TRT] ConvReluFusion: Fusing inception_3b/1x1 with inception_3b/relu_1x1 [TRT] ConvReluFusion: Fusing inception_3b/3x3_reduce with inception_3b/relu_3x3_reduce [TRT] ConvReluFusion: Fusing inception_3b/3x3 with inception_3b/relu_3x3 [TRT] ConvReluFusion: Fusing inception_3b/5x5_reduce with inception_3b/relu_5x5_reduce [TRT] ConvReluFusion: Fusing inception_3b/5x5 with inception_3b/relu_5x5 [TRT] ConvReluFusion: Fusing inception_3b/pool_proj with inception_3b/relu_pool_proj [TRT] ConvReluFusion: Fusing inception_4a/1x1 with inception_4a/relu_1x1 [TRT] ConvReluFusion: Fusing inception_4a/3x3_reduce with inception_4a/relu_3x3_reduce [TRT] ConvReluFusion: Fusing inception_4a/3x3 with inception_4a/relu_3x3 [TRT] ConvReluFusion: Fusing inception_4a/5x5_reduce with inception_4a/relu_5x5_reduce [TRT] ConvReluFusion: Fusing inception_4a/5x5 with inception_4a/relu_5x5 [TRT] ConvReluFusion: Fusing inception_4a/pool_proj with inception_4a/relu_pool_proj [TRT] ConvReluFusion: Fusing inception_4b/1x1 with inception_4b/relu_1x1 [TRT] ConvReluFusion: Fusing inception_4b/3x3_reduce with inception_4b/relu_3x3_reduce [TRT] ConvReluFusion: Fusing inception_4b/3x3 with inception_4b/relu_3x3 [TRT] ConvReluFusion: Fusing inception_4b/5x5_reduce with inception_4b/relu_5x5_reduce [TRT] ConvReluFusion: Fusing inception_4b/5x5 with inception_4b/relu_5x5 [TRT] ConvReluFusion: Fusing inception_4b/pool_proj with inception_4b/relu_pool_proj [TRT] ConvReluFusion: Fusing inception_4c/1x1 with inception_4c/relu_1x1 [TRT] ConvReluFusion: Fusing inception_4c/3x3_reduce with inception_4c/relu_3x3_reduce [TRT] ConvReluFusion: Fusing inception_4c/3x3 with inception_4c/relu_3x3 [TRT] ConvReluFusion: Fusing inception_4c/5x5_reduce with inception_4c/relu_5x5_reduce [TRT] ConvReluFusion: Fusing inception_4c/5x5 with inception_4c/relu_5x5 [TRT] ConvReluFusion: Fusing inception_4c/pool_proj with inception_4c/relu_pool_proj [TRT] ConvReluFusion: Fusing inception_4d/1x1 with inception_4d/relu_1x1 [TRT] ConvReluFusion: Fusing inception_4d/3x3_reduce with inception_4d/relu_3x3_reduce [TRT] ConvReluFusion: Fusing inception_4d/3x3 with inception_4d/relu_3x3 [TRT] ConvReluFusion: Fusing inception_4d/5x5_reduce with inception_4d/relu_5x5_reduce [TRT] ConvReluFusion: Fusing inception_4d/5x5 with inception_4d/relu_5x5 [TRT] ConvReluFusion: Fusing inception_4d/pool_proj with inception_4d/relu_pool_proj [TRT] ConvReluFusion: Fusing inception_4e/1x1 with inception_4e/relu_1x1 [TRT] ConvReluFusion: Fusing inception_4e/3x3_reduce with inception_4e/relu_3x3_reduce [TRT] ConvReluFusion: Fusing inception_4e/3x3 with inception_4e/relu_3x3 [TRT] ConvReluFusion: Fusing inception_4e/5x5_reduce with inception_4e/relu_5x5_reduce [TRT] ConvReluFusion: Fusing inception_4e/5x5 with inception_4e/relu_5x5 [TRT] ConvReluFusion: Fusing inception_4e/pool_proj with inception_4e/relu_pool_proj [TRT] ConvReluFusion: Fusing inception_5a/1x1 with inception_5a/relu_1x1 [TRT] ConvReluFusion: Fusing inception_5a/3x3_reduce with inception_5a/relu_3x3_reduce [TRT] ConvReluFusion: Fusing inception_5a/3x3 with inception_5a/relu_3x3 [TRT] ConvReluFusion: Fusing inception_5a/5x5_reduce with inception_5a/relu_5x5_reduce [TRT] ConvReluFusion: Fusing inception_5a/5x5 with inception_5a/relu_5x5 [TRT] ConvReluFusion: Fusing inception_5a/pool_proj with inception_5a/relu_pool_proj [TRT] ConvReluFusion: Fusing inception_5b/1x1 with inception_5b/relu_1x1 [TRT] ConvReluFusion: Fusing inception_5b/3x3_reduce with inception_5b/relu_3x3_reduce [TRT] ConvReluFusion: Fusing inception_5b/3x3 with inception_5b/relu_3x3 [TRT] ConvReluFusion: Fusing inception_5b/5x5_reduce with inception_5b/relu_5x5_reduce [TRT] ConvReluFusion: Fusing inception_5b/5x5 with inception_5b/relu_5x5 [TRT] ConvReluFusion: Fusing inception_5b/pool_proj with inception_5b/relu_pool_proj [TRT] After vertical fusions: 84 layers [TRT] After dupe layer removal: 84 layers [TRT] After final dead-layer removal: 84 layers [TRT] Merging layers: inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce [TRT] Merging layers: inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce [TRT] Merging layers: inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce [TRT] Merging layers: inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce [TRT] Merging layers: inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce [TRT] Merging layers: inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce [TRT] Merging layers: inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce [TRT] Merging layers: inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce [TRT] Merging layers: inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce [TRT] After tensor merging: 66 layers [TRT] Eliminating concatenation inception_3a/output [TRT] Generating copy for inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce to inception_3a/output because input is not movable. [TRT] Retargeting inception_3a/3x3 to inception_3a/output [TRT] Retargeting inception_3a/5x5 to inception_3a/output [TRT] Retargeting inception_3a/pool_proj to inception_3a/output [TRT] Eliminating concatenation inception_3b/output [TRT] Generating copy for inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce to inception_3b/output because input is not movable. [TRT] Retargeting inception_3b/3x3 to inception_3b/output [TRT] Retargeting inception_3b/5x5 to inception_3b/output [TRT] Retargeting inception_3b/pool_proj to inception_3b/output [TRT] Eliminating concatenation inception_4a/output [TRT] Generating copy for inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce to inception_4a/output because input is not movable. [TRT] Retargeting inception_4a/3x3 to inception_4a/output [TRT] Retargeting inception_4a/5x5 to inception_4a/output [TRT] Retargeting inception_4a/pool_proj to inception_4a/output [TRT] Eliminating concatenation inception_4b/output [TRT] Generating copy for inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce to inception_4b/output because input is not movable. [TRT] Retargeting inception_4b/3x3 to inception_4b/output [TRT] Retargeting inception_4b/5x5 to inception_4b/output [TRT] Retargeting inception_4b/pool_proj to inception_4b/output [TRT] Eliminating concatenation inception_4c/output [TRT] Generating copy for inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce to inception_4c/output because input is not movable. [TRT] Retargeting inception_4c/3x3 to inception_4c/output [TRT] Retargeting inception_4c/5x5 to inception_4c/output [TRT] Retargeting inception_4c/pool_proj to inception_4c/output [TRT] Eliminating concatenation inception_4d/output [TRT] Generating copy for inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce to inception_4d/output because input is not movable. [TRT] Retargeting inception_4d/3x3 to inception_4d/output [TRT] Retargeting inception_4d/5x5 to inception_4d/output [TRT] Retargeting inception_4d/pool_proj to inception_4d/output [TRT] Eliminating concatenation inception_4e/output [TRT] Generating copy for inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce to inception_4e/output because input is not movable. [TRT] Retargeting inception_4e/3x3 to inception_4e/output [TRT] Retargeting inception_4e/5x5 to inception_4e/output [TRT] Retargeting inception_4e/pool_proj to inception_4e/output [TRT] Eliminating concatenation inception_5a/output [TRT] Generating copy for inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce to inception_5a/output because input is not movable. [TRT] Retargeting inception_5a/3x3 to inception_5a/output [TRT] Retargeting inception_5a/5x5 to inception_5a/output [TRT] Retargeting inception_5a/pool_proj to inception_5a/output [TRT] Eliminating concatenation inception_5b/output [TRT] Generating copy for inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce to inception_5b/output because input is not movable. [TRT] Retargeting inception_5b/3x3 to inception_5b/output [TRT] Retargeting inception_5b/5x5 to inception_5b/output [TRT] Retargeting inception_5b/pool_proj to inception_5b/output [TRT] After concat removal: 66 layers [TRT] Graph construction and optimization completed in 0.0479605 seconds. [TRT] ---------- Layers Running on DLA ---------- [TRT] ---------- Layers Running on GPU ---------- [TRT] [GpuLayer] conv1/7x7_s2 + conv1/relu_7x7 [TRT] [GpuLayer] pool1/3x3_s2 [TRT] [GpuLayer] pool1/norm1 [TRT] [GpuLayer] conv2/3x3_reduce + conv2/relu_3x3_reduce [TRT] [GpuLayer] conv2/3x3 + conv2/relu_3x3 [TRT] [GpuLayer] conv2/norm2 [TRT] [GpuLayer] pool2/3x3_s2 [TRT] [GpuLayer] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce [TRT] [GpuLayer] inception_3a/3x3 + inception_3a/relu_3x3 [TRT] [GpuLayer] inception_3a/5x5 + inception_3a/relu_5x5 [TRT] [GpuLayer] inception_3a/pool [TRT] [GpuLayer] inception_3a/pool_proj + inception_3a/relu_pool_proj [TRT] [GpuLayer] inception_3a/1x1 copy [TRT] [GpuLayer] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce [TRT] [GpuLayer] inception_3b/3x3 + inception_3b/relu_3x3 [TRT] [GpuLayer] inception_3b/5x5 + inception_3b/relu_5x5 [TRT] [GpuLayer] inception_3b/pool [TRT] [GpuLayer] inception_3b/pool_proj + inception_3b/relu_pool_proj [TRT] [GpuLayer] inception_3b/1x1 copy [TRT] [GpuLayer] pool3/3x3_s2 [TRT] [GpuLayer] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce [TRT] [GpuLayer] inception_4a/3x3 + inception_4a/relu_3x3 [TRT] [GpuLayer] inception_4a/5x5 + inception_4a/relu_5x5 [TRT] [GpuLayer] inception_4a/pool [TRT] [GpuLayer] inception_4a/pool_proj + inception_4a/relu_pool_proj [TRT] [GpuLayer] inception_4a/1x1 copy [TRT] [GpuLayer] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce [TRT] [GpuLayer] inception_4b/3x3 + inception_4b/relu_3x3 [TRT] [GpuLayer] inception_4b/5x5 + inception_4b/relu_5x5 [TRT] [GpuLayer] inception_4b/pool [TRT] [GpuLayer] inception_4b/pool_proj + inception_4b/relu_pool_proj [TRT] [GpuLayer] inception_4b/1x1 copy [TRT] [GpuLayer] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce [TRT] [GpuLayer] inception_4c/3x3 + inception_4c/relu_3x3 [TRT] [GpuLayer] inception_4c/5x5 + inception_4c/relu_5x5 [TRT] [GpuLayer] inception_4c/pool [TRT] [GpuLayer] inception_4c/pool_proj + inception_4c/relu_pool_proj [TRT] [GpuLayer] inception_4c/1x1 copy [TRT] [GpuLayer] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce [TRT] [GpuLayer] inception_4d/3x3 + inception_4d/relu_3x3 [TRT] [GpuLayer] inception_4d/5x5 + inception_4d/relu_5x5 [TRT] [GpuLayer] inception_4d/pool [TRT] [GpuLayer] inception_4d/pool_proj + inception_4d/relu_pool_proj [TRT] [GpuLayer] inception_4d/1x1 copy [TRT] [GpuLayer] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce [TRT] [GpuLayer] inception_4e/3x3 + inception_4e/relu_3x3 [TRT] [GpuLayer] inception_4e/5x5 + inception_4e/relu_5x5 [TRT] [GpuLayer] inception_4e/pool [TRT] [GpuLayer] inception_4e/pool_proj + inception_4e/relu_pool_proj [TRT] [GpuLayer] inception_4e/1x1 copy [TRT] [GpuLayer] pool4/3x3_s2 [TRT] [GpuLayer] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce [TRT] [GpuLayer] inception_5a/3x3 + inception_5a/relu_3x3 [TRT] [GpuLayer] inception_5a/5x5 + inception_5a/relu_5x5 [TRT] [GpuLayer] inception_5a/pool [TRT] [GpuLayer] inception_5a/pool_proj + inception_5a/relu_pool_proj [TRT] [GpuLayer] inception_5a/1x1 copy [TRT] [GpuLayer] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce [TRT] [GpuLayer] inception_5b/3x3 + inception_5b/relu_3x3 [TRT] [GpuLayer] inception_5b/5x5 + inception_5b/relu_5x5 [TRT] [GpuLayer] inception_5b/pool [TRT] [GpuLayer] inception_5b/pool_proj + inception_5b/relu_pool_proj [TRT] [GpuLayer] inception_5b/1x1 copy [TRT] [GpuLayer] pool5/7x7_s1 [TRT] [GpuLayer] loss3/classifier [TRT] [GpuLayer] prob [TRT] Using cublas a tactic source [TRT] [MemUsageChange] Init cuBLAS/cuBLASLt: CPU +158, GPU +77, now: CPU 453, GPU 3468 (MiB) [TRT] Using cuDNN as a tactic source [TRT] [MemUsageChange] Init cuDNN: CPU +241, GPU -6, now: CPU 694, GPU 3462 (MiB) [TRT] Detected invalid timing cache, setup a local cache instead [TRT] Constructing optimization profile number 0 [1/1]. [TRT] *************** Autotuning Reformat:Float(150528,50176,224,1) -> Float(150528,1,672,3) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 13.2983 [TRT] Tactic: 0 Time: 0.449739 [TRT] Fastest Tactic: 0 Time: 0.449739 [TRT] *************** Autotuning Reformat:Float(150528,50176,224,1) -> Half(150528,50176,224,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.378229 [TRT] Tactic: 0 Time: 0.312552 [TRT] Fastest Tactic: 0 Time: 0.312552 [TRT] *************** Autotuning Reformat:Float(150528,50176,224,1) -> Half(100352,50176:2,224,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.842682 [TRT] Tactic: 0 Time: 0.661171 [TRT] Fastest Tactic: 0 Time: 0.661171 [TRT] *************** Autotuning format combination: Float(150528,50176,224,1) -> Float(802816,12544,112,1) *************** [TRT] --------------- Timing Runner: conv1/7x7_s2 + conv1/relu_7x7 (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: conv1/7x7_s2 + conv1/relu_7x7 (FusedConvActConvolution) [TRT] FusedConvActConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: conv1/7x7_s2 + conv1/relu_7x7 (CudnnConvolution) [TRT] Tactic: 0 Time: 13.9327 [TRT] Tactic: 1 Time: 5.76826 [TRT] Tactic: 2 Time: 13.519 [TRT] Tactic: 5 Time: 100.695 [TRT] Fastest Tactic: 1 Time: 5.76826 [TRT] --------------- Timing Runner: conv1/7x7_s2 + conv1/relu_7x7 (CaskConvolution) [TRT] conv1/7x7_s2 + conv1/relu_7x7 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 1.90409 [TRT] conv1/7x7_s2 + conv1/relu_7x7 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v0 Tactic: 1754984623894446479 [TRT] Tactic: 1754984623894446479 Time: 1.94372 [TRT] conv1/7x7_s2 + conv1/relu_7x7 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v0 Tactic: 3611739942397549984 [TRT] Tactic: 3611739942397549984 Time: 3.03833 [TRT] conv1/7x7_s2 + conv1/relu_7x7 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v1 Tactic: 4337000649858996379 [TRT] Tactic: 4337000649858996379 Time: 1.49729 [TRT] conv1/7x7_s2 + conv1/relu_7x7 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 2.9376 [TRT] conv1/7x7_s2 + conv1/relu_7x7 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 1.47719 [TRT] conv1/7x7_s2 + conv1/relu_7x7 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v1 Tactic: -9137461792520977713 [TRT] Tactic: -9137461792520977713 Time: 3.0299 [TRT] conv1/7x7_s2 + conv1/relu_7x7 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 2.99792 [TRT] conv1/7x7_s2 + conv1/relu_7x7 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v0 Tactic: -8133971918129952780 [TRT] Tactic: -8133971918129952780 Time: 1.55412 [TRT] conv1/7x7_s2 + conv1/relu_7x7 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v1 Tactic: -6092040395344634144 [TRT] Tactic: -6092040395344634144 Time: 1.95776 [TRT] conv1/7x7_s2 + conv1/relu_7x7 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 1.93971 [TRT] conv1/7x7_s2 + conv1/relu_7x7 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 1.54307 [TRT] Fastest Tactic: 6645123197870846056 Time: 1.47719 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 6645123197870846056 [TRT] *************** Autotuning format combination: Float(150528,1,672,3) -> Float(802816,1,7168,64) *************** [TRT] --------------- Timing Runner: conv1/7x7_s2 + conv1/relu_7x7 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: conv1/7x7_s2 + conv1/relu_7x7 (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] *************** Autotuning format combination: Half(150528,50176,224,1) -> Half(802816,12544,112,1) *************** [TRT] --------------- Timing Runner: conv1/7x7_s2 + conv1/relu_7x7 (CudnnConvolution) [TRT] Tactic: 0 Time: 3.76794 [TRT] Tactic: 1 Time: 3.16664 [TRT] Tactic: 2 Time: 4.8144 [TRT] Tactic: 5 Time: 101.187 [TRT] Fastest Tactic: 1 Time: 3.16664 [TRT] --------------- Timing Runner: conv1/7x7_s2 + conv1/relu_7x7 (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 1 [TRT] *************** Autotuning format combination: Half(100352,50176:2,224,1) -> Half(401408,12544:2,112,1) *************** [TRT] --------------- Timing Runner: conv1/7x7_s2 + conv1/relu_7x7 (FusedConvActConvolution) [TRT] FusedConvActConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: conv1/7x7_s2 + conv1/relu_7x7 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: conv1/7x7_s2 + conv1/relu_7x7 (CaskConvolution) [TRT] conv1/7x7_s2 + conv1/relu_7x7 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 1.32307 [TRT] conv1/7x7_s2 + conv1/relu_7x7 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_large_nn_v1 Tactic: 3650389455493082349 [TRT] Tactic: 3650389455493082349 Time: 1.36781 [TRT] conv1/7x7_s2 + conv1/relu_7x7 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 1.02799 [TRT] conv1/7x7_s2 + conv1/relu_7x7 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_large_nn_v1 Tactic: -6490690591794140522 [TRT] Tactic: -6490690591794140522 Time: 1.03617 [TRT] conv1/7x7_s2 + conv1/relu_7x7 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_large_nn_v1 Tactic: -4686027666808657977 [TRT] Tactic: -4686027666808657977 Time: 2.05935 [TRT] conv1/7x7_s2 + conv1/relu_7x7 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 2.01857 [TRT] Fastest Tactic: 7205456024582378848 Time: 1.02799 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 7205456024582378848 [TRT] *************** Autotuning Reformat:Float(802816,12544,112,1) -> Half(802816,12544,112,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.998567 [TRT] Tactic: 0 Time: 0.813359 [TRT] Fastest Tactic: 0 Time: 0.813359 [TRT] *************** Autotuning Reformat:Float(802816,12544,112,1) -> Half(401408,12544:2,112,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.2019 [TRT] Tactic: 0 Time: 0.65026 [TRT] Fastest Tactic: 0 Time: 0.65026 [TRT] *************** Autotuning Reformat:Float(802816,1,7168,64) -> Float(802816,12544,112,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.941641 [TRT] Tactic: 0 Time: 2.70865 [TRT] Fastest Tactic: 1002 Time: 0.941641 [TRT] *************** Autotuning Reformat:Float(802816,1,7168,64) -> Half(802816,12544,112,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.746511 [TRT] Tactic: 0 Time: 2.65633 [TRT] Fastest Tactic: 1002 Time: 0.746511 [TRT] *************** Autotuning Reformat:Float(802816,1,7168,64) -> Half(401408,12544:2,112,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.03552 [TRT] Tactic: 0 Time: 2.84682 [TRT] Fastest Tactic: 1002 Time: 1.03552 [TRT] *************** Autotuning Reformat:Half(802816,12544,112,1) -> Float(802816,12544,112,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.00909 [TRT] Tactic: 0 Time: 0.693437 [TRT] Fastest Tactic: 0 Time: 0.693437 [TRT] *************** Autotuning Reformat:Half(802816,12544,112,1) -> Half(401408,12544:2,112,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.677578 [TRT] Tactic: 0 Time: 0.643541 [TRT] Fastest Tactic: 0 Time: 0.643541 [TRT] *************** Autotuning Reformat:Half(401408,12544:2,112,1) -> Float(802816,12544,112,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.972188 [TRT] Tactic: 0 Time: 0.562943 [TRT] Fastest Tactic: 0 Time: 0.562943 [TRT] *************** Autotuning Reformat:Half(401408,12544:2,112,1) -> Half(802816,12544,112,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.52253 [TRT] Tactic: 0 Time: 0.549115 [TRT] Fastest Tactic: 0 Time: 0.549115 [TRT] *************** Autotuning format combination: Float(802816,12544,112,1) -> Float(200704,3136,56,1) *************** [TRT] --------------- Timing Runner: pool1/3x3_s2 (TiledPooling) [TRT] Tactic: 257 Time: 0.980573 [TRT] Tactic: 65793 Time: 0.825912 [TRT] Tactic: 131329 Time: 1.1676 [TRT] Tactic: 196865 Time: 1.23766 [TRT] Tactic: 262401 Time: 1.01424 [TRT] Tactic: 327937 Time: 1.06453 [TRT] Tactic: 393473 Time: 1.11919 [TRT] Tactic: 459009 Time: 0.641823 [TRT] Tactic: 524545 Time: 0.532943 [TRT] Tactic: 590081 Time: 0.762578 [TRT] Tactic: 655617 Time: 0.779297 [TRT] Tactic: 721153 Time: 0.630885 [TRT] Tactic: 786689 Time: 0.665756 [TRT] Tactic: 852225 Time: 0.714687 [TRT] Tactic: 917761 Time: 0.570338 [TRT] Tactic: 983297 Time: 0.447682 [TRT] Tactic: 1048833 Time: 0.643281 [TRT] Tactic: 1114369 Time: 0.626276 [TRT] Tactic: 1179905 Time: 0.517187 [TRT] Tactic: 1245441 Time: 0.556562 [TRT] Tactic: 1310977 Time: 0.593177 [TRT] Tactic: 1376513 Time: 0.549714 [TRT] Tactic: 1442049 Time: 0.424766 [TRT] Tactic: 1507585 Time: 0.537109 [TRT] Tactic: 1573121 Time: 0.52724 [TRT] Tactic: 1638657 Time: 0.432968 [TRT] Tactic: 1704193 Time: 0.46862 [TRT] Tactic: 1769729 Time: 0.500808 [TRT] Tactic: 1835265 Time: 0.540442 [TRT] Tactic: 1900801 Time: 0.420625 [TRT] Tactic: 1966337 Time: 0.496536 [TRT] Tactic: 2031873 Time: 0.474948 [TRT] Tactic: 2097409 Time: 0.395521 [TRT] Tactic: 2162945 Time: 0.438359 [TRT] Tactic: 2228481 Time: 0.456641 [TRT] Tactic: 2294017 Time: 0.542526 [TRT] Tactic: 2359553 Time: 0.417943 [TRT] Tactic: 2425089 Time: 0.49125 [TRT] Tactic: 2490625 Time: 0.444401 [TRT] Tactic: 2556161 Time: 0.379245 [TRT] Tactic: 2621697 Time: 0.389531 [TRT] Tactic: 2687233 Time: 0.437708 [TRT] Tactic: 6947073 Time: 0.340807 [TRT] Fastest Tactic: 6947073 Time: 0.340807 [TRT] --------------- Timing Runner: pool1/3x3_s2 (CudnnPooling) [TRT] Tactic: -1 Time: 1.26607 [TRT] Fastest Tactic: -1 Time: 1.26607 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: TiledPooling Tactic: 6947073 [TRT] *************** Autotuning format combination: Half(802816,12544,112,1) -> Half(200704,3136,56,1) *************** [TRT] --------------- Timing Runner: pool1/3x3_s2 (TiledPooling) [TRT] TiledPooling has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: pool1/3x3_s2 (CudnnPooling) [TRT] Tactic: -1 Time: 1.29714 [TRT] Fastest Tactic: -1 Time: 1.29714 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnPooling Tactic: -1 [TRT] *************** Autotuning format combination: Half(401408,12544:2,112,1) -> Half(100352,3136:2,56,1) *************** [TRT] --------------- Timing Runner: pool1/3x3_s2 (TiledPooling) [TRT] Tactic: 257 Time: 0.507578 [TRT] Tactic: 65793 Time: 0.430885 [TRT] Tactic: 131329 Time: 0.59901 [TRT] Tactic: 196865 Time: 0.655104 [TRT] Tactic: 262401 Time: 0.523749 [TRT] Tactic: 327937 Time: 0.573099 [TRT] Tactic: 393473 Time: 0.580261 [TRT] Tactic: 459009 Time: 0.353307 [TRT] Tactic: 524545 Time: 0.290729 [TRT] Tactic: 590081 Time: 0.398645 [TRT] Tactic: 655617 Time: 0.420312 [TRT] Tactic: 721153 Time: 0.341511 [TRT] Tactic: 786689 Time: 0.362292 [TRT] Tactic: 852225 Time: 0.381615 [TRT] Tactic: 917761 Time: 0.286484 [TRT] Tactic: 983297 Time: 0.23776 [TRT] Tactic: 1048833 Time: 0.341042 [TRT] Tactic: 1114369 Time: 0.335261 [TRT] Tactic: 1179905 Time: 0.274896 [TRT] Tactic: 1245441 Time: 0.285886 [TRT] Tactic: 1310977 Time: 0.318255 [TRT] Tactic: 1376513 Time: 0.277422 [TRT] Tactic: 1442049 Time: 0.212916 [TRT] Tactic: 1507585 Time: 0.291667 [TRT] Tactic: 1573121 Time: 0.295443 [TRT] Tactic: 1638657 Time: 0.237318 [TRT] Tactic: 1704193 Time: 0.26013 [TRT] Tactic: 1769729 Time: 0.271953 [TRT] Tactic: 1835265 Time: 0.274375 [TRT] Tactic: 1900801 Time: 0.214011 [TRT] Tactic: 1966337 Time: 0.279271 [TRT] Tactic: 2031873 Time: 0.263255 [TRT] Tactic: 2097409 Time: 0.236744 [TRT] Tactic: 2162945 Time: 0.236407 [TRT] Tactic: 2228481 Time: 0.250573 [TRT] Tactic: 2294017 Time: 0.273724 [TRT] Tactic: 2359553 Time: 0.213438 [TRT] Tactic: 2425089 Time: 0.287995 [TRT] Tactic: 2490625 Time: 0.277942 [TRT] Tactic: 2556161 Time: 0.228516 [TRT] Tactic: 2621697 Time: 0.206849 [TRT] Tactic: 2687233 Time: 0.265391 [TRT] Tactic: 6947073 Time: 0.203958 [TRT] Fastest Tactic: 6947073 Time: 0.203958 [TRT] --------------- Timing Runner: pool1/3x3_s2 (CudaPooling) [TRT] Tactic: -3 Time: 0.368984 [TRT] Fastest Tactic: -3 Time: 0.368984 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: TiledPooling Tactic: 6947073 [TRT] *************** Autotuning Reformat:Float(200704,3136,56,1) -> Float(200704,1,3584,64) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.208776 [TRT] Tactic: 0 Time: 0.331901 [TRT] Fastest Tactic: 1002 Time: 0.208776 [TRT] *************** Autotuning Reformat:Float(200704,3136,56,1) -> Half(200704,3136,56,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.283307 [TRT] Tactic: 0 Time: 0.210443 [TRT] Fastest Tactic: 0 Time: 0.210443 [TRT] *************** Autotuning Reformat:Float(200704,3136,56,1) -> Half(100352,3136:2,56,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.30289 [TRT] Tactic: 0 Time: 0.167787 [TRT] Fastest Tactic: 0 Time: 0.167787 [TRT] *************** Autotuning Reformat:Float(200704,1,3584,64) -> Float(200704,3136,56,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.256068 [TRT] Tactic: 0 Time: 0.682396 [TRT] Fastest Tactic: 1002 Time: 0.256068 [TRT] *************** Autotuning Reformat:Float(200704,1,3584,64) -> Half(200704,3136,56,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.192812 [TRT] Tactic: 0 Time: 0.670651 [TRT] Fastest Tactic: 1002 Time: 0.192812 [TRT] *************** Autotuning Reformat:Float(200704,1,3584,64) -> Half(100352,3136:2,56,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.258959 [TRT] Tactic: 0 Time: 0.712604 [TRT] Fastest Tactic: 1002 Time: 0.258959 [TRT] *************** Autotuning Reformat:Half(200704,3136,56,1) -> Float(200704,3136,56,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.285573 [TRT] Tactic: 0 Time: 0.177995 [TRT] Fastest Tactic: 0 Time: 0.177995 [TRT] *************** Autotuning Reformat:Half(200704,3136,56,1) -> Float(200704,1,3584,64) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.181823 [TRT] Tactic: 0 Time: 0.305052 [TRT] Fastest Tactic: 1002 Time: 0.181823 [TRT] *************** Autotuning Reformat:Half(200704,3136,56,1) -> Half(100352,3136:2,56,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.180599 [TRT] Tactic: 0 Time: 0.166041 [TRT] Fastest Tactic: 0 Time: 0.166041 [TRT] *************** Autotuning Reformat:Half(100352,3136:2,56,1) -> Float(200704,3136,56,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.244766 [TRT] Tactic: 0 Time: 0.144713 [TRT] Fastest Tactic: 0 Time: 0.144713 [TRT] *************** Autotuning Reformat:Half(100352,3136:2,56,1) -> Float(200704,1,3584,64) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.185339 [TRT] Tactic: 0 Time: 0.35263 [TRT] Fastest Tactic: 1002 Time: 0.185339 [TRT] *************** Autotuning Reformat:Half(100352,3136:2,56,1) -> Half(200704,3136,56,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.379115 [TRT] Tactic: 0 Time: 0.142552 [TRT] Fastest Tactic: 0 Time: 0.142552 [TRT] *************** Autotuning format combination: Float(200704,3136,56,1) -> Float(200704,3136,56,1) *************** [TRT] --------------- Timing Runner: pool1/norm1 (CudaLRN) [TRT] CudaLRN has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: pool1/norm1 (CudnnLRN) [TRT] Tactic: 0 Time: 0.14612 [TRT] Fastest Tactic: 0 Time: 0.14612 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnLRN Tactic: 0 [TRT] *************** Autotuning format combination: Float(200704,1,3584,64) -> Float(200704,1,3584,64) *************** [TRT] --------------- Timing Runner: pool1/norm1 (CudaLRN) [TRT] CudaLRN has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: pool1/norm1 (CudnnLRN) [TRT] CudnnLRN has no valid tactics for this config, skipping [TRT] *************** Autotuning format combination: Half(200704,3136,56,1) -> Half(200704,3136,56,1) *************** [TRT] --------------- Timing Runner: pool1/norm1 (CudaLRN) [TRT] CudaLRN has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: pool1/norm1 (CudnnLRN) [TRT] Tactic: 0 Time: 0.170754 [TRT] Fastest Tactic: 0 Time: 0.170754 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnLRN Tactic: 0 [TRT] *************** Autotuning format combination: Half(100352,3136:2,56,1) -> Half(100352,3136:2,56,1) *************** [TRT] --------------- Timing Runner: pool1/norm1 (CudaLRN) [TRT] Tactic: 0 Time: 0.0738805 [TRT] Fastest Tactic: 0 Time: 0.0738805 [TRT] --------------- Timing Runner: pool1/norm1 (CudnnLRN) [TRT] CudnnLRN has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudaLRN Tactic: 0 [TRT] *************** Autotuning Reformat:Float(200704,3136,56,1) -> Float(200704,1,3584,64) *************** [TRT] *************** Autotuning Reformat:Float(200704,3136,56,1) -> Half(200704,3136,56,1) *************** [TRT] *************** Autotuning Reformat:Float(200704,3136,56,1) -> Half(100352,3136:2,56,1) *************** [TRT] *************** Autotuning Reformat:Float(200704,1,3584,64) -> Float(200704,3136,56,1) *************** [TRT] *************** Autotuning Reformat:Float(200704,1,3584,64) -> Half(200704,3136,56,1) *************** [TRT] *************** Autotuning Reformat:Float(200704,1,3584,64) -> Half(100352,3136:2,56,1) *************** [TRT] *************** Autotuning Reformat:Half(200704,3136,56,1) -> Float(200704,3136,56,1) *************** [TRT] *************** Autotuning Reformat:Half(200704,3136,56,1) -> Float(200704,1,3584,64) *************** [TRT] *************** Autotuning Reformat:Half(200704,3136,56,1) -> Half(100352,3136:2,56,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,3136:2,56,1) -> Float(200704,3136,56,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,3136:2,56,1) -> Float(200704,1,3584,64) *************** [TRT] *************** Autotuning Reformat:Half(100352,3136:2,56,1) -> Half(200704,3136,56,1) *************** [TRT] *************** Autotuning format combination: Float(200704,3136,56,1) -> Float(200704,3136,56,1) *************** [TRT] --------------- Timing Runner: conv2/3x3_reduce + conv2/relu_3x3_reduce (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: conv2/3x3_reduce + conv2/relu_3x3_reduce (FusedConvActConvolution) [TRT] FusedConvActConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: conv2/3x3_reduce + conv2/relu_3x3_reduce (CudnnConvolution) [TRT] Tactic: 0 Time: 0.573359 [TRT] Tactic: 1 Time: 0.518958 [TRT] Tactic: 2 Time: 0.834244 [TRT] Tactic: 4 skipped. Scratch requested: 140591104, available: 33554432 [TRT] Tactic: 5 Time: 1.2688 [TRT] Some tactics do not have sufficient workspace memory to run. Increasing workspace size may increase performance, please check verbose output. [TRT] Fastest Tactic: 1 Time: 0.518958 [TRT] --------------- Timing Runner: conv2/3x3_reduce + conv2/relu_3x3_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: conv2/3x3_reduce + conv2/relu_3x3_reduce (CaskConvolution) [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.279063 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v0 Tactic: 1698681053543049347 [TRT] Tactic: 1698681053543049347 Time: 0.251433 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.39974 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.209896 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.41375 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v1 Tactic: 5326823351883942011 [TRT] Tactic: 5326823351883942011 Time: 0.388463 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v0 Tactic: 5500448035057547314 [TRT] Tactic: 5500448035057547314 Time: 0.222995 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.213827 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.280651 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.419922 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v1 Tactic: -6576203419454146580 [TRT] Tactic: -6576203419454146580 Time: 0.25875 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.289037 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.264661 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.236719 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.233777 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.400495 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v0 Tactic: -377491875521947884 [TRT] Tactic: -377491875521947884 Time: 0.410052 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v1 Tactic: -37215280111360163 [TRT] Tactic: -37215280111360163 Time: 0.205078 [TRT] Fastest Tactic: -37215280111360163 Time: 0.205078 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -37215280111360163 [TRT] *************** Autotuning format combination: Float(200704,1,3584,64) -> Float(200704,1,3584,64) *************** [TRT] --------------- Timing Runner: conv2/3x3_reduce + conv2/relu_3x3_reduce (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: conv2/3x3_reduce + conv2/relu_3x3_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: conv2/3x3_reduce + conv2/relu_3x3_reduce (CaskConvolution) [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 3886731678879822788 [TRT] Tactic: 3886731678879822788 Time: 0.238646 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 6629944304117643200 [TRT] Tactic: 6629944304117643200 Time: 0.604063 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.60974 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.242213 [TRT] Fastest Tactic: 3886731678879822788 Time: 0.238646 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 3886731678879822788 [TRT] *************** Autotuning format combination: Half(200704,3136,56,1) -> Half(200704,3136,56,1) *************** [TRT] --------------- Timing Runner: conv2/3x3_reduce + conv2/relu_3x3_reduce (CudnnConvolution) [TRT] Tactic: 0 Time: 0.539505 [TRT] Tactic: 1 Time: 0.536693 [TRT] Tactic: 2 Time: 0.795052 [TRT] Tactic: 4 skipped. Scratch requested: 140591104, available: 33554432 [TRT] Tactic: 5 Time: 1.20341 [TRT] Fastest Tactic: 1 Time: 0.536693 [TRT] --------------- Timing Runner: conv2/3x3_reduce + conv2/relu_3x3_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: conv2/3x3_reduce + conv2/relu_3x3_reduce (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 1 [TRT] *************** Autotuning format combination: Half(100352,3136:2,56,1) -> Half(200704,3136,56,1) *************** [TRT] --------------- Timing Runner: conv2/3x3_reduce + conv2/relu_3x3_reduce (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] *************** Autotuning format combination: Half(100352,3136:2,56,1) -> Half(100352,3136:2,56,1) *************** [TRT] --------------- Timing Runner: conv2/3x3_reduce + conv2/relu_3x3_reduce (FusedConvActConvolution) [TRT] FusedConvActConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: conv2/3x3_reduce + conv2/relu_3x3_reduce (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: conv2/3x3_reduce + conv2/relu_3x3_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: conv2/3x3_reduce + conv2/relu_3x3_reduce (CaskConvolution) [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_interior_nn_v1 Tactic: 3066127711859985668 [TRT] Tactic: 3066127711859985668 Time: 0.151484 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.162291 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.156511 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.124349 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] Tactic: 8163473458334948789 Time: 0.121198 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.227839 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.229089 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.122396 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_interior_nn_v1 Tactic: -1716393687483585322 [TRT] Tactic: -1716393687483585322 Time: 0.227656 [TRT] Fastest Tactic: 8163473458334948789 Time: 0.121198 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 8163473458334948789 [TRT] *************** Autotuning Reformat:Float(200704,3136,56,1) -> Float(200704,1,3584,64) *************** [TRT] *************** Autotuning Reformat:Float(200704,3136,56,1) -> Half(200704,3136,56,1) *************** [TRT] *************** Autotuning Reformat:Float(200704,3136,56,1) -> Half(100352,3136:2,56,1) *************** [TRT] *************** Autotuning Reformat:Float(200704,1,3584,64) -> Float(200704,3136,56,1) *************** [TRT] *************** Autotuning Reformat:Float(200704,1,3584,64) -> Half(200704,3136,56,1) *************** [TRT] *************** Autotuning Reformat:Float(200704,1,3584,64) -> Half(100352,3136:2,56,1) *************** [TRT] *************** Autotuning Reformat:Half(200704,3136,56,1) -> Float(200704,3136,56,1) *************** [TRT] *************** Autotuning Reformat:Half(200704,3136,56,1) -> Float(200704,1,3584,64) *************** [TRT] *************** Autotuning Reformat:Half(200704,3136,56,1) -> Half(100352,3136:2,56,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,3136:2,56,1) -> Float(200704,3136,56,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,3136:2,56,1) -> Float(200704,1,3584,64) *************** [TRT] *************** Autotuning Reformat:Half(100352,3136:2,56,1) -> Half(200704,3136,56,1) *************** [TRT] *************** Autotuning format combination: Float(200704,3136,56,1) -> Float(602112,3136,56,1) *************** [TRT] --------------- Timing Runner: conv2/3x3 + conv2/relu_3x3 (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: conv2/3x3 + conv2/relu_3x3 (FusedConvActConvolution) [TRT] FusedConvActConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: conv2/3x3 + conv2/relu_3x3 (CudnnConvolution) [TRT] Tactic: 0 Time: 8.78536 [TRT] Tactic: 1 Time: 4.35461 [TRT] Tactic: 2 Time: 8.1276 [TRT] Tactic: 4 skipped. Scratch requested: 417841152, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 55705600, available: 33554432 [TRT] Tactic: 6 Time: 3.47776 [TRT] Fastest Tactic: 6 Time: 3.47776 [TRT] --------------- Timing Runner: conv2/3x3 + conv2/relu_3x3 (CaskConvolution) [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 4.59901 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v0 Tactic: 1754984623894446479 [TRT] Tactic: 1754984623894446479 Time: 5.12784 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v0 Tactic: 3611739942397549984 [TRT] Tactic: 3611739942397549984 Time: 4.97781 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148n_nt_v1 Tactic: 3827454225649558724 [TRT] Tactic: 3827454225649558724 Time: 4.14789 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v1 Tactic: 4337000649858996379 [TRT] Tactic: 4337000649858996379 Time: 3.77406 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 4.97896 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 3.64583 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 4.9181 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 5921334924264294896 [TRT] Tactic: 5921334924264294896 Time: 2.98299 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 3.71878 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 4.67279 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148t_nt_v1 Tactic: 7852627285308570038 [TRT] Tactic: 7852627285308570038 Time: 4.16604 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v1 Tactic: -9137461792520977713 [TRT] Tactic: -9137461792520977713 Time: 4.98979 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148n_nt_v0 Tactic: -8776506421218919509 [TRT] Tactic: -8776506421218919509 Time: 4.08253 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 5.01927 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v0 Tactic: -8133971918129952780 [TRT] Tactic: -8133971918129952780 Time: 4.16034 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v1 Tactic: -6092040395344634144 [TRT] Tactic: -6092040395344634144 Time: 4.75451 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 5.10219 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 4.20409 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148t_nt_v0 Tactic: -2318106587342035239 [TRT] Tactic: -2318106587342035239 Time: 4.13305 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_mobile_relu_tile148t_nt_v0 Tactic: -1343271414618805657 [TRT] Tactic: -1343271414618805657 Time: 2.71852 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 3.993 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 3.88841 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 4.86383 [TRT] Fastest Tactic: -1343271414618805657 Time: 2.71852 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -1343271414618805657 [TRT] *************** Autotuning format combination: Float(200704,1,3584,64) -> Float(602112,1,10752,192) *************** [TRT] --------------- Timing Runner: conv2/3x3 + conv2/relu_3x3 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: conv2/3x3 + conv2/relu_3x3 (CaskConvolution) [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 5.15755 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 3.63102 [TRT] Fastest Tactic: -7394439838318485025 Time: 3.63102 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(200704,3136,56,1) -> Half(602112,3136,56,1) *************** [TRT] --------------- Timing Runner: conv2/3x3 + conv2/relu_3x3 (CudnnConvolution) [TRT] Tactic: 0 Time: 8.69969 [TRT] Tactic: 1 Time: 7.38482 [TRT] Tactic: 2 Time: 7.66672 [TRT] Tactic: 4 skipped. Scratch requested: 417841152, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 55705600, available: 33554432 [TRT] Tactic: 6 Time: 4.51424 [TRT] Fastest Tactic: 6 Time: 4.51424 [TRT] --------------- Timing Runner: conv2/3x3 + conv2/relu_3x3 (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 6 [TRT] *************** Autotuning format combination: Half(100352,3136:2,56,1) -> Half(301056,3136:2,56,1) *************** [TRT] --------------- Timing Runner: conv2/3x3 + conv2/relu_3x3 (FusedConvActConvolution) [TRT] FusedConvActConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: conv2/3x3 + conv2/relu_3x3 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: conv2/3x3 + conv2/relu_3x3 (CaskConvolution) [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 2.41992 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_large_nn_v1 Tactic: 3650389455493082349 [TRT] Tactic: 3650389455493082349 Time: 2.50688 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_winograd_fp16x2_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 4772821744921268633 [TRT] Tactic: 4772821744921268633 Time: 1.6388 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 2.20349 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 1.90526 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_large_nn_v1 Tactic: -6490690591794140522 [TRT] Tactic: -6490690591794140522 Time: 1.94383 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_large_nn_v1 Tactic: -4686027666808657977 [TRT] Tactic: -4686027666808657977 Time: 2.5443 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 2.44445 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 2.52789 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 1.85297 [TRT] Fastest Tactic: 4772821744921268633 Time: 1.6388 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 4772821744921268633 [TRT] *************** Autotuning Reformat:Float(602112,3136,56,1) -> Float(602112,1,10752,192) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.612525 [TRT] Tactic: 0 Time: 1.05922 [TRT] Fastest Tactic: 1002 Time: 0.612525 [TRT] *************** Autotuning Reformat:Float(602112,3136,56,1) -> Half(602112,3136,56,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.832942 [TRT] Tactic: 0 Time: 0.612735 [TRT] Fastest Tactic: 0 Time: 0.612735 [TRT] *************** Autotuning Reformat:Float(602112,3136,56,1) -> Half(301056,3136:2,56,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.899844 [TRT] Tactic: 0 Time: 0.489818 [TRT] Fastest Tactic: 0 Time: 0.489818 [TRT] *************** Autotuning Reformat:Float(602112,1,10752,192) -> Float(602112,3136,56,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.713516 [TRT] Tactic: 0 Time: 2.25971 [TRT] Fastest Tactic: 1002 Time: 0.713516 [TRT] *************** Autotuning Reformat:Float(602112,1,10752,192) -> Half(602112,3136,56,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.561615 [TRT] Tactic: 0 Time: 2.20078 [TRT] Fastest Tactic: 1002 Time: 0.561615 [TRT] *************** Autotuning Reformat:Float(602112,1,10752,192) -> Half(301056,3136:2,56,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.770052 [TRT] Tactic: 0 Time: 2.34188 [TRT] Fastest Tactic: 1002 Time: 0.770052 [TRT] *************** Autotuning Reformat:Half(602112,3136,56,1) -> Float(602112,3136,56,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.839818 [TRT] Tactic: 0 Time: 0.522058 [TRT] Fastest Tactic: 0 Time: 0.522058 [TRT] *************** Autotuning Reformat:Half(602112,3136,56,1) -> Float(602112,1,10752,192) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.527266 [TRT] Tactic: 0 Time: 0.958021 [TRT] Fastest Tactic: 1002 Time: 0.527266 [TRT] *************** Autotuning Reformat:Half(602112,3136,56,1) -> Half(301056,3136:2,56,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.514844 [TRT] Tactic: 0 Time: 0.484479 [TRT] Fastest Tactic: 0 Time: 0.484479 [TRT] *************** Autotuning Reformat:Half(301056,3136:2,56,1) -> Float(602112,3136,56,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.708411 [TRT] Tactic: 0 Time: 0.422552 [TRT] Fastest Tactic: 0 Time: 0.422552 [TRT] *************** Autotuning Reformat:Half(301056,3136:2,56,1) -> Float(602112,1,10752,192) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.536041 [TRT] Tactic: 0 Time: 1.04406 [TRT] Fastest Tactic: 1002 Time: 0.536041 [TRT] *************** Autotuning Reformat:Half(301056,3136:2,56,1) -> Half(602112,3136,56,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.09643 [TRT] Tactic: 0 Time: 0.413854 [TRT] Fastest Tactic: 0 Time: 0.413854 [TRT] *************** Autotuning format combination: Float(602112,3136,56,1) -> Float(602112,3136,56,1) *************** [TRT] --------------- Timing Runner: conv2/norm2 (CudaLRN) [TRT] CudaLRN has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: conv2/norm2 (CudnnLRN) [TRT] Tactic: 0 Time: 0.435833 [TRT] Fastest Tactic: 0 Time: 0.435833 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnLRN Tactic: 0 [TRT] *************** Autotuning format combination: Float(602112,1,10752,192) -> Float(602112,1,10752,192) *************** [TRT] --------------- Timing Runner: conv2/norm2 (CudaLRN) [TRT] CudaLRN has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: conv2/norm2 (CudnnLRN) [TRT] CudnnLRN has no valid tactics for this config, skipping [TRT] *************** Autotuning format combination: Half(602112,3136,56,1) -> Half(602112,3136,56,1) *************** [TRT] --------------- Timing Runner: conv2/norm2 (CudaLRN) [TRT] CudaLRN has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: conv2/norm2 (CudnnLRN) [TRT] Tactic: 0 Time: 0.496224 [TRT] Fastest Tactic: 0 Time: 0.496224 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnLRN Tactic: 0 [TRT] *************** Autotuning format combination: Half(301056,3136:2,56,1) -> Half(301056,3136:2,56,1) *************** [TRT] --------------- Timing Runner: conv2/norm2 (CudaLRN) [TRT] Tactic: 0 Time: 0.20724 [TRT] Fastest Tactic: 0 Time: 0.20724 [TRT] --------------- Timing Runner: conv2/norm2 (CudnnLRN) [TRT] CudnnLRN has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudaLRN Tactic: 0 [TRT] *************** Autotuning Reformat:Float(602112,3136,56,1) -> Half(602112,3136,56,1) *************** [TRT] *************** Autotuning Reformat:Float(602112,3136,56,1) -> Half(301056,3136:2,56,1) *************** [TRT] *************** Autotuning Reformat:Float(602112,1,10752,192) -> Float(602112,3136,56,1) *************** [TRT] *************** Autotuning Reformat:Float(602112,1,10752,192) -> Half(602112,3136,56,1) *************** [TRT] *************** Autotuning Reformat:Float(602112,1,10752,192) -> Half(301056,3136:2,56,1) *************** [TRT] *************** Autotuning Reformat:Half(602112,3136,56,1) -> Float(602112,3136,56,1) *************** [TRT] *************** Autotuning Reformat:Half(602112,3136,56,1) -> Half(301056,3136:2,56,1) *************** [TRT] *************** Autotuning Reformat:Half(301056,3136:2,56,1) -> Float(602112,3136,56,1) *************** [TRT] *************** Autotuning Reformat:Half(301056,3136:2,56,1) -> Half(602112,3136,56,1) *************** [TRT] *************** Autotuning format combination: Float(602112,3136,56,1) -> Float(150528,784,28,1) *************** [TRT] --------------- Timing Runner: pool2/3x3_s2 (TiledPooling) [TRT] Tactic: 257 Time: 0.726041 [TRT] Tactic: 65793 Time: 0.701224 [TRT] Tactic: 131329 Time: 0.867109 [TRT] Tactic: 196865 Time: 1.84828 [TRT] Tactic: 262401 Time: 1.46917 [TRT] Tactic: 327937 Time: 0.780938 [TRT] Tactic: 393473 Time: 0.821901 [TRT] Tactic: 459009 Time: 0.470703 [TRT] Tactic: 524545 Time: 0.45401 [TRT] Tactic: 590081 Time: 0.567657 [TRT] Tactic: 655617 Time: 1.10789 [TRT] Tactic: 721153 Time: 0.870286 [TRT] Tactic: 786689 Time: 0.479948 [TRT] Tactic: 852225 Time: 0.514974 [TRT] Tactic: 917761 Time: 0.402655 [TRT] Tactic: 983297 Time: 0.382786 [TRT] Tactic: 1048833 Time: 0.461354 [TRT] Tactic: 1114369 Time: 0.847578 [TRT] Tactic: 1179905 Time: 0.682734 [TRT] Tactic: 1245441 Time: 0.386875 [TRT] Tactic: 1310977 Time: 0.41414 [TRT] Tactic: 1376513 Time: 0.390651 [TRT] Tactic: 1442049 Time: 0.367031 [TRT] Tactic: 1507585 Time: 0.400651 [TRT] Tactic: 1573121 Time: 0.697708 [TRT] Tactic: 1638657 Time: 0.576485 [TRT] Tactic: 1704193 Time: 0.32349 [TRT] Tactic: 1769729 Time: 0.358308 [TRT] Tactic: 1835265 Time: 0.389244 [TRT] Tactic: 1900801 Time: 0.361172 [TRT] Tactic: 1966337 Time: 0.370703 [TRT] Tactic: 2031873 Time: 0.635494 [TRT] Tactic: 2097409 Time: 0.52185 [TRT] Tactic: 2162945 Time: 0.303906 [TRT] Tactic: 2228481 Time: 0.324584 [TRT] Tactic: 2294017 Time: 0.38961 [TRT] Tactic: 2359553 Time: 0.362553 [TRT] Tactic: 2425089 Time: 0.351589 [TRT] Tactic: 2490625 Time: 0.552396 [TRT] Tactic: 2556161 Time: 0.471016 [TRT] Tactic: 2621697 Time: 0.269114 [TRT] Tactic: 2687233 Time: 0.286875 [TRT] Tactic: 6947073 Time: 0.252682 [TRT] Fastest Tactic: 6947073 Time: 0.252682 [TRT] --------------- Timing Runner: pool2/3x3_s2 (CudnnPooling) [TRT] Tactic: -1 Time: 0.991562 [TRT] Fastest Tactic: -1 Time: 0.991562 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: TiledPooling Tactic: 6947073 [TRT] *************** Autotuning format combination: Half(602112,3136,56,1) -> Half(150528,784,28,1) *************** [TRT] --------------- Timing Runner: pool2/3x3_s2 (TiledPooling) [TRT] TiledPooling has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: pool2/3x3_s2 (CudnnPooling) [TRT] Tactic: -1 Time: 1.02508 [TRT] Fastest Tactic: -1 Time: 1.02508 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnPooling Tactic: -1 [TRT] *************** Autotuning format combination: Half(301056,3136:2,56,1) -> Half(75264,784:2,28,1) *************** [TRT] --------------- Timing Runner: pool2/3x3_s2 (TiledPooling) [TRT] Tactic: 257 Time: 0.379974 [TRT] Tactic: 65793 Time: 0.368177 [TRT] Tactic: 131329 Time: 0.447839 [TRT] Tactic: 196865 Time: 0.981485 [TRT] Tactic: 262401 Time: 0.774974 [TRT] Tactic: 327937 Time: 0.425313 [TRT] Tactic: 393473 Time: 0.429141 [TRT] Tactic: 459009 Time: 0.257968 [TRT] Tactic: 524545 Time: 0.24849 [TRT] Tactic: 590081 Time: 0.299844 [TRT] Tactic: 655617 Time: 0.605417 [TRT] Tactic: 721153 Time: 0.487734 [TRT] Tactic: 786689 Time: 0.259531 [TRT] Tactic: 852225 Time: 0.27862 [TRT] Tactic: 917761 Time: 0.209532 [TRT] Tactic: 983297 Time: 0.202136 [TRT] Tactic: 1048833 Time: 0.251015 [TRT] Tactic: 1114369 Time: 0.462266 [TRT] Tactic: 1179905 Time: 0.38625 [TRT] Tactic: 1245441 Time: 0.199219 [TRT] Tactic: 1310977 Time: 0.224896 [TRT] Tactic: 1376513 Time: 0.197891 [TRT] Tactic: 1442049 Time: 0.189505 [TRT] Tactic: 1507585 Time: 0.213021 [TRT] Tactic: 1573121 Time: 0.395833 [TRT] Tactic: 1638657 Time: 0.328463 [TRT] Tactic: 1704193 Time: 0.174479 [TRT] Tactic: 1769729 Time: 0.190729 [TRT] Tactic: 1835265 Time: 0.19737 [TRT] Tactic: 1900801 Time: 0.188515 [TRT] Tactic: 1966337 Time: 0.208542 [TRT] Tactic: 2031873 Time: 0.379323 [TRT] Tactic: 2097409 Time: 0.318854 [TRT] Tactic: 2162945 Time: 0.166249 [TRT] Tactic: 2228481 Time: 0.174896 [TRT] Tactic: 2294017 Time: 0.199219 [TRT] Tactic: 2359553 Time: 0.186042 [TRT] Tactic: 2425089 Time: 0.170026 [TRT] Tactic: 2490625 Time: 0.322838 [TRT] Tactic: 2556161 Time: 0.285833 [TRT] Tactic: 2621697 Time: 0.136797 [TRT] Tactic: 2687233 Time: 0.155105 [TRT] Tactic: 6947073 Time: 0.151693 [TRT] Fastest Tactic: 2621697 Time: 0.136797 [TRT] --------------- Timing Runner: pool2/3x3_s2 (CudaPooling) [TRT] Tactic: -3 Time: 0.28138 [TRT] Fastest Tactic: -3 Time: 0.28138 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: TiledPooling Tactic: 2621697 [TRT] *************** Autotuning Reformat:Float(150528,784,28,1) -> Float(150528,1,5376,192) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.166328 [TRT] Tactic: 0 Time: 0.246094 [TRT] Fastest Tactic: 1002 Time: 0.166328 [TRT] *************** Autotuning Reformat:Float(150528,784,28,1) -> Half(150528,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.24612 [TRT] Tactic: 0 Time: 0.160755 [TRT] Fastest Tactic: 0 Time: 0.160755 [TRT] *************** Autotuning Reformat:Float(150528,784,28,1) -> Half(75264,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.224636 [TRT] Tactic: 0 Time: 0.127708 [TRT] Fastest Tactic: 0 Time: 0.127708 [TRT] *************** Autotuning Reformat:Half(150528,784,28,1) -> Float(150528,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.248854 [TRT] Tactic: 0 Time: 0.134791 [TRT] Fastest Tactic: 0 Time: 0.134791 [TRT] *************** Autotuning Reformat:Half(150528,784,28,1) -> Float(150528,1,5376,192) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.141381 [TRT] Tactic: 0 Time: 0.230521 [TRT] Fastest Tactic: 1002 Time: 0.141381 [TRT] *************** Autotuning Reformat:Half(150528,784,28,1) -> Half(75264,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.152005 [TRT] Tactic: 0 Time: 0.126172 [TRT] Fastest Tactic: 0 Time: 0.126172 [TRT] *************** Autotuning Reformat:Half(75264,784:2,28,1) -> Float(150528,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.186328 [TRT] Tactic: 0 Time: 0.109818 [TRT] Fastest Tactic: 0 Time: 0.109818 [TRT] *************** Autotuning Reformat:Half(75264,784:2,28,1) -> Float(150528,1,5376,192) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.141823 [TRT] Tactic: 0 Time: 0.266797 [TRT] Fastest Tactic: 1002 Time: 0.141823 [TRT] *************** Autotuning Reformat:Half(75264,784:2,28,1) -> Half(150528,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.277344 [TRT] Tactic: 0 Time: 0.107344 [TRT] Fastest Tactic: 0 Time: 0.107344 [TRT] *************** Autotuning Reformat:Float(150528,784,28,1) -> Float(150528,1,5376,192) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.166718 [TRT] Tactic: 0 Time: 0.244218 [TRT] Fastest Tactic: 1002 Time: 0.166718 [TRT] *************** Autotuning Reformat:Float(150528,784,28,1) -> Half(150528,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.245182 [TRT] Tactic: 0 Time: 0.160417 [TRT] Fastest Tactic: 0 Time: 0.160417 [TRT] *************** Autotuning Reformat:Float(150528,784,28,1) -> Half(75264,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.224454 [TRT] Tactic: 0 Time: 0.127917 [TRT] Fastest Tactic: 0 Time: 0.127917 [TRT] *************** Autotuning Reformat:Float(150528,1,5376,192) -> Float(150528,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.197109 [TRT] Tactic: 0 Time: 0.242291 [TRT] Fastest Tactic: 1002 Time: 0.197109 [TRT] *************** Autotuning Reformat:Float(150528,1,5376,192) -> Half(150528,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.148776 [TRT] Tactic: 0 Time: 0.238828 [TRT] Fastest Tactic: 1002 Time: 0.148776 [TRT] *************** Autotuning Reformat:Float(150528,1,5376,192) -> Half(75264,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.197005 [TRT] Tactic: 0 Time: 0.271406 [TRT] Fastest Tactic: 1002 Time: 0.197005 [TRT] *************** Autotuning Reformat:Half(150528,784,28,1) -> Float(150528,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.24802 [TRT] Tactic: 0 Time: 0.13461 [TRT] Fastest Tactic: 0 Time: 0.13461 [TRT] *************** Autotuning Reformat:Half(150528,784,28,1) -> Float(150528,1,5376,192) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.14073 [TRT] Tactic: 0 Time: 0.229609 [TRT] Fastest Tactic: 1002 Time: 0.14073 [TRT] *************** Autotuning Reformat:Half(150528,784,28,1) -> Half(75264,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.15112 [TRT] Tactic: 0 Time: 0.125651 [TRT] Fastest Tactic: 0 Time: 0.125651 [TRT] *************** Autotuning Reformat:Half(75264,784:2,28,1) -> Float(150528,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.186485 [TRT] Tactic: 0 Time: 0.110494 [TRT] Fastest Tactic: 0 Time: 0.110494 [TRT] *************** Autotuning Reformat:Half(75264,784:2,28,1) -> Float(150528,1,5376,192) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.142291 [TRT] Tactic: 0 Time: 0.266198 [TRT] Fastest Tactic: 1002 Time: 0.142291 [TRT] *************** Autotuning Reformat:Half(75264,784:2,28,1) -> Half(150528,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.276511 [TRT] Tactic: 0 Time: 0.107474 [TRT] Fastest Tactic: 0 Time: 0.107474 [TRT] *************** Autotuning format combination: Float(150528,784,28,1) -> Float(137984,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce (FusedConvActConvolution) [TRT] FusedConvActConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce (CudnnConvolution) [TRT] Tactic: 0 Time: 0.848256 [TRT] Tactic: 1 Time: 0.851979 [TRT] Tactic: 2 Time: 0.940495 [TRT] Tactic: 4 skipped. Scratch requested: 295931904, available: 33554432 [TRT] Tactic: 5 Time: 2.88411 [TRT] Fastest Tactic: 0 Time: 0.848256 [TRT] --------------- Timing Runner: inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce (CaskConvolution) [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.505469 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v0 Tactic: 1698681053543049347 [TRT] Tactic: 1698681053543049347 Time: 0.479323 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.517838 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.391719 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.524739 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v1 Tactic: 5326823351883942011 [TRT] Tactic: 5326823351883942011 Time: 0.49586 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v0 Tactic: 5500448035057547314 [TRT] Tactic: 5500448035057547314 Time: 0.430885 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.402734 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.541718 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.526979 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v1 Tactic: -6576203419454146580 [TRT] Tactic: -6576203419454146580 Time: 0.467057 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.567343 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.486328 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.443333 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.44151 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.510287 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v0 Tactic: -377491875521947884 [TRT] Tactic: -377491875521947884 Time: 0.514948 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v1 Tactic: -37215280111360163 [TRT] Tactic: -37215280111360163 Time: 0.385364 [TRT] Fastest Tactic: -37215280111360163 Time: 0.385364 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -37215280111360163 [TRT] *************** Autotuning format combination: Float(150528,1,5376,192) -> Float(137984,1,4928,176) *************** [TRT] --------------- Timing Runner: inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce (CaskConvolution) [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 3886731678879822788 [TRT] Tactic: 3886731678879822788 Time: 0.411537 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 6629944304117643200 [TRT] Tactic: 6629944304117643200 Time: 0.816589 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.823334 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.412526 [TRT] Fastest Tactic: 3886731678879822788 Time: 0.411537 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 3886731678879822788 [TRT] *************** Autotuning format combination: Half(150528,784,28,1) -> Half(137984,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce (CudnnConvolution) [TRT] Tactic: 0 Time: 0.838463 [TRT] Tactic: 1 Time: 0.729245 [TRT] Tactic: 2 Time: 0.897213 [TRT] Tactic: 4 skipped. Scratch requested: 295931904, available: 33554432 [TRT] Tactic: 5 Time: 2.83003 [TRT] Fastest Tactic: 1 Time: 0.729245 [TRT] --------------- Timing Runner: inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 1 [TRT] *************** Autotuning format combination: Half(75264,784:2,28,1) -> Half(137984,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] *************** Autotuning format combination: Half(75264,784:2,28,1) -> Half(68992,784:2,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce (FusedConvActConvolution) [TRT] FusedConvActConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce (CaskConvolution) [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_interior_nn_v1 Tactic: 3066127711859985668 [TRT] Tactic: 3066127711859985668 Time: 0.253359 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.284037 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.264895 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.217187 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] Tactic: 8163473458334948789 Time: 0.210026 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.274948 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.278099 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.213151 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_interior_nn_v1 Tactic: -1716393687483585322 [TRT] Tactic: -1716393687483585322 Time: 0.270391 [TRT] Fastest Tactic: 8163473458334948789 Time: 0.210026 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 8163473458334948789 [TRT] *************** Autotuning Reformat:Float(137984,784,28,1) -> Float(137984,1,4928,176) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.173229 [TRT] Tactic: 0 Time: 0.225808 [TRT] Fastest Tactic: 1002 Time: 0.173229 [TRT] *************** Autotuning Reformat:Float(137984,784,28,1) -> Half(137984,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.225938 [TRT] Tactic: 0 Time: 0.146875 [TRT] Fastest Tactic: 0 Time: 0.146875 [TRT] *************** Autotuning Reformat:Float(137984,784,28,1) -> Half(68992,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.208437 [TRT] Tactic: 0 Time: 0.117005 [TRT] Fastest Tactic: 0 Time: 0.117005 [TRT] *************** Autotuning Reformat:Float(137984,1,4928,176) -> Float(137984,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.197318 [TRT] Tactic: 0 Time: 0.224869 [TRT] Fastest Tactic: 1002 Time: 0.197318 [TRT] *************** Autotuning Reformat:Float(137984,1,4928,176) -> Half(137984,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.151666 [TRT] Tactic: 0 Time: 0.222318 [TRT] Fastest Tactic: 1002 Time: 0.151666 [TRT] *************** Autotuning Reformat:Float(137984,1,4928,176) -> Half(68992,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.18263 [TRT] Tactic: 0 Time: 0.253802 [TRT] Fastest Tactic: 1002 Time: 0.18263 [TRT] *************** Autotuning Reformat:Half(137984,784,28,1) -> Float(137984,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.228828 [TRT] Tactic: 0 Time: 0.124792 [TRT] Fastest Tactic: 0 Time: 0.124792 [TRT] *************** Autotuning Reformat:Half(137984,784,28,1) -> Float(137984,1,4928,176) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.143359 [TRT] Tactic: 0 Time: 0.21099 [TRT] Fastest Tactic: 1002 Time: 0.143359 [TRT] *************** Autotuning Reformat:Half(137984,784,28,1) -> Half(68992,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.138229 [TRT] Tactic: 0 Time: 0.116042 [TRT] Fastest Tactic: 0 Time: 0.116042 [TRT] *************** Autotuning Reformat:Half(68992,784:2,28,1) -> Float(137984,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.180183 [TRT] Tactic: 0 Time: 0.101224 [TRT] Fastest Tactic: 0 Time: 0.101224 [TRT] *************** Autotuning Reformat:Half(68992,784:2,28,1) -> Float(137984,1,4928,176) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.144818 [TRT] Tactic: 0 Time: 0.244792 [TRT] Fastest Tactic: 1002 Time: 0.144818 [TRT] *************** Autotuning Reformat:Half(68992,784:2,28,1) -> Half(137984,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.26349 [TRT] Tactic: 0 Time: 0.0988025 [TRT] Fastest Tactic: 0 Time: 0.0988025 [TRT] *************** Autotuning Reformat:Float(137984,784,28,1) -> Float(137984,1,4928,176) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.11513 [TRT] Tactic: 0 Time: 0.125052 [TRT] Fastest Tactic: 1002 Time: 0.11513 [TRT] *************** Autotuning Reformat:Float(137984,784,28,1) -> Half(137984,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.14888 [TRT] Tactic: 0 Time: 0.108776 [TRT] Fastest Tactic: 0 Time: 0.108776 [TRT] *************** Autotuning Reformat:Float(137984,784,28,1) -> Half(68992,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.117292 [TRT] Tactic: 0 Time: 0.129271 [TRT] Fastest Tactic: 1002 Time: 0.117292 [TRT] *************** Autotuning Reformat:Float(137984,1,4928,176) -> Float(137984,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.132032 [TRT] Tactic: 0 Time: 0.126016 [TRT] Fastest Tactic: 0 Time: 0.126016 [TRT] *************** Autotuning Reformat:Float(137984,1,4928,176) -> Half(137984,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.078099 [TRT] Tactic: 0 Time: 0.124349 [TRT] Fastest Tactic: 1002 Time: 0.078099 [TRT] *************** Autotuning Reformat:Float(137984,1,4928,176) -> Half(68992,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.10388 [TRT] Tactic: 0 Time: 0.141355 [TRT] Fastest Tactic: 1002 Time: 0.10388 [TRT] *************** Autotuning Reformat:Half(137984,784,28,1) -> Float(137984,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.150286 [TRT] Tactic: 0 Time: 0.0702085 [TRT] Fastest Tactic: 0 Time: 0.0702085 [TRT] *************** Autotuning Reformat:Half(137984,784,28,1) -> Float(137984,1,4928,176) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.07388 [TRT] Tactic: 0 Time: 0.118047 [TRT] Fastest Tactic: 1002 Time: 0.07388 [TRT] *************** Autotuning Reformat:Half(137984,784,28,1) -> Half(68992,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.08086 [TRT] Tactic: 0 Time: 0.065781 [TRT] Fastest Tactic: 0 Time: 0.065781 [TRT] *************** Autotuning Reformat:Half(68992,784:2,28,1) -> Float(137984,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0988545 [TRT] Tactic: 0 Time: 0.057708 [TRT] Fastest Tactic: 0 Time: 0.057708 [TRT] *************** Autotuning Reformat:Half(68992,784:2,28,1) -> Float(137984,1,4928,176) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0748445 [TRT] Tactic: 0 Time: 0.136615 [TRT] Fastest Tactic: 1002 Time: 0.0748445 [TRT] *************** Autotuning Reformat:Half(68992,784:2,28,1) -> Half(137984,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.142604 [TRT] Tactic: 0 Time: 0.11461 [TRT] Fastest Tactic: 0 Time: 0.11461 [TRT] *************** Autotuning format combination: Float(137984,784,28,1) -> Float(200704,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/3x3 + inception_3a/relu_3x3 (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3a/3x3 + inception_3a/relu_3x3 (FusedConvActConvolution) [TRT] Tactic: 524287 Time: 1.07526 [TRT] Tactic: 720895 Time: 1.15195 [TRT] Tactic: 983039 Time: 1.02714 [TRT] Tactic: 1048575 Time: 1.25924 [TRT] Tactic: 1703935 Time: 1.25214 [TRT] Tactic: 1769471 Time: 1.25672 [TRT] Tactic: 1966079 Time: 1.10909 [TRT] Tactic: 2031615 Time: 1.16281 [TRT] Tactic: 2228223 Time: 1.35505 [TRT] Tactic: 2424831 Time: 1.53326 [TRT] Tactic: 2621439 Time: 1.57839 [TRT] Tactic: 2752511 Time: 1.11987 [TRT] Tactic: 2818047 Time: 1.63016 [TRT] Tactic: 2883583 Time: 1.1649 [TRT] Tactic: 3014655 Time: 1.18844 [TRT] Tactic: 3145727 Time: 1.11221 [TRT] Tactic: 3473407 Time: 1.18961 [TRT] Tactic: 3604479 Time: 1.16508 [TRT] Tactic: 3735551 Time: 1.40112 [TRT] Tactic: 4390911 Time: 1.08039 [TRT] Tactic: 5046271 Time: 1.06383 [TRT] Tactic: 5963775 Time: 0.991771 [TRT] Tactic: 6160383 Time: 1.05214 [TRT] Tactic: 6488063 Time: 1.09292 [TRT] Tactic: 6881279 Time: 1.12633 [TRT] Tactic: 7274495 Time: 1.59727 [TRT] Tactic: 7864319 Time: 1.39156 [TRT] Tactic: 7995391 Time: 1.17687 [TRT] Tactic: 8585215 Time: 1.09456 [TRT] Tactic: 8847359 Time: 1.23115 [TRT] Tactic: 8978431 Time: 1.02357 [TRT] Tactic: 9043967 Time: 1.08669 [TRT] Tactic: 9175039 Time: 1.17013 [TRT] Tactic: 9502719 Time: 1.11435 [TRT] Tactic: 9830399 Time: 1.07224 [TRT] Tactic: 9961471 Time: 1.35427 [TRT] Tactic: 10027007 Time: 1.08419 [TRT] Tactic: 10092543 Time: 1.09539 [TRT] Tactic: 10289151 Time: 1.11026 [TRT] Tactic: 10485759 Time: 1.09831 [TRT] Tactic: 10682367 Time: 1.54992 [TRT] Tactic: 10813439 Time: 1.16646 [TRT] Fastest Tactic: 5963775 Time: 0.991771 [TRT] --------------- Timing Runner: inception_3a/3x3 + inception_3a/relu_3x3 (CudnnConvolution) [TRT] Tactic: 0 Time: 1.86414 [TRT] Tactic: 1 Time: 1.47964 [TRT] Tactic: 2 Time: 1.93122 [TRT] Tactic: 4 skipped. Scratch requested: 108232704, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 54452224, available: 33554432 [TRT] Tactic: 6 Time: 0.969244 [TRT] Fastest Tactic: 6 Time: 0.969244 [TRT] --------------- Timing Runner: inception_3a/3x3 + inception_3a/relu_3x3 (CaskConvolution) [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 1.4325 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v0 Tactic: 1754984623894446479 [TRT] Tactic: 1754984623894446479 Time: 1.62953 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v0 Tactic: 3611739942397549984 [TRT] Tactic: 3611739942397549984 Time: 1.18594 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148n_nt_v1 Tactic: 3827454225649558724 [TRT] Tactic: 3827454225649558724 Time: 1.10146 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v1 Tactic: 4337000649858996379 [TRT] Tactic: 4337000649858996379 Time: 1.19471 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 1.16286 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 1.12135 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 1.1437 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 5921334924264294896 [TRT] Tactic: 5921334924264294896 Time: 0.830703 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 1.16409 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 1.43675 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148t_nt_v1 Tactic: 7852627285308570038 [TRT] Tactic: 7852627285308570038 Time: 1.10401 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v1 Tactic: -9137461792520977713 [TRT] Tactic: -9137461792520977713 Time: 1.16956 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148n_nt_v0 Tactic: -8776506421218919509 [TRT] Tactic: -8776506421218919509 Time: 1.07323 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 1.1837 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v0 Tactic: -8133971918129952780 [TRT] Tactic: -8133971918129952780 Time: 1.29466 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v1 Tactic: -6092040395344634144 [TRT] Tactic: -6092040395344634144 Time: 1.49258 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 1.62583 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 1.27518 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148t_nt_v0 Tactic: -2318106587342035239 [TRT] Tactic: -2318106587342035239 Time: 1.10167 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_mobile_relu_tile148t_nt_v0 Tactic: -1343271414618805657 [TRT] Tactic: -1343271414618805657 Time: 0.769714 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 1.26984 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 1.21852 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 1.11732 [TRT] Fastest Tactic: -1343271414618805657 Time: 0.769714 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -1343271414618805657 [TRT] *************** Autotuning format combination: Float(137984,1,4928,176) -> Float(200704,1,7168,256) *************** [TRT] --------------- Timing Runner: inception_3a/3x3 + inception_3a/relu_3x3 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3a/3x3 + inception_3a/relu_3x3 (CaskConvolution) [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 1.44674 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 1.09078 [TRT] Fastest Tactic: -7394439838318485025 Time: 1.09078 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(137984,784,28,1) -> Half(200704,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/3x3 + inception_3a/relu_3x3 (CudnnConvolution) [TRT] Tactic: 0 Time: 2.44708 [TRT] Tactic: 1 Time: 2.45221 [TRT] Tactic: 2 Time: 1.8575 [TRT] Tactic: 4 skipped. Scratch requested: 108232704, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 54452224, available: 33554432 [TRT] Tactic: 6 Time: 3.54904 [TRT] Fastest Tactic: 2 Time: 1.8575 [TRT] --------------- Timing Runner: inception_3a/3x3 + inception_3a/relu_3x3 (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 2 [TRT] *************** Autotuning format combination: Half(68992,784:2,28,1) -> Half(100352,784:2,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/3x3 + inception_3a/relu_3x3 (FusedConvActConvolution) [TRT] Tactic: 524287 Time: 0.600703 [TRT] Tactic: 720895 Time: 0.720182 [TRT] Tactic: 983039 Time: 0.601511 [TRT] Tactic: 1048575 Time: 0.757995 [TRT] Tactic: 1703935 Time: 0.78513 [TRT] Tactic: 1769471 Time: 4.31128 [TRT] Tactic: 1966079 Time: 0.700547 [TRT] Tactic: 2031615 Time: 0.652552 [TRT] Tactic: 2228223 Time: 0.807318 [TRT] Tactic: 2424831 Time: 1.24391 [TRT] Tactic: 2621439 Time: 0.931927 [TRT] Tactic: 2752511 Time: 0.662604 [TRT] Tactic: 2818047 Time: 0.936198 [TRT] Tactic: 2883583 Time: 0.74349 [TRT] Tactic: 3014655 Time: 0.747058 [TRT] Tactic: 3145727 Time: 0.706249 [TRT] Tactic: 3473407 Time: 0.708958 [TRT] Tactic: 3604479 Time: 0.734323 [TRT] Tactic: 3735551 Time: 0.632734 [TRT] Tactic: 4390911 Time: 0.666979 [TRT] Tactic: 5046271 Time: 0.623932 [TRT] Tactic: 5963775 Time: 0.568828 [TRT] Tactic: 6160383 Time: 0.630702 [TRT] Tactic: 6488063 Time: 0.659999 [TRT] Tactic: 6881279 Time: 0.61336 [TRT] Tactic: 7274495 Time: 0.862265 [TRT] Tactic: 7864319 Time: 0.840235 [TRT] Tactic: 7995391 Time: 0.721303 [TRT] Tactic: 8585215 Time: 0.652604 [TRT] Tactic: 8847359 Time: 0.753047 [TRT] Tactic: 8978431 Time: 0.605989 [TRT] Tactic: 9043967 Time: 0.660521 [TRT] Tactic: 9175039 Time: 0.732162 [TRT] Tactic: 9502719 Time: 0.661693 [TRT] Tactic: 9830399 Time: 0.610912 [TRT] Tactic: 9961471 Time: 0.82651 [TRT] Tactic: 10027007 Time: 0.649844 [TRT] Tactic: 10092543 Time: 0.667214 [TRT] Tactic: 10289151 Time: 0.693933 [TRT] Tactic: 10485759 Time: 0.660547 [TRT] Tactic: 10682367 Time: 0.908828 [TRT] Tactic: 10813439 Time: 0.747292 [TRT] Fastest Tactic: 5963775 Time: 0.568828 [TRT] --------------- Timing Runner: inception_3a/3x3 + inception_3a/relu_3x3 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3a/3x3 + inception_3a/relu_3x3 (CaskConvolution) [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.752838 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_large_nn_v1 Tactic: 3650389455493082349 [TRT] Tactic: 3650389455493082349 Time: 0.771824 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_winograd_fp16x2_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 4772821744921268633 [TRT] Tactic: 4772821744921268633 Time: 0.458046 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.674375 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.589219 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_large_nn_v1 Tactic: -6490690591794140522 [TRT] Tactic: -6490690591794140522 Time: 0.595886 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_large_nn_v1 Tactic: -4686027666808657977 [TRT] Tactic: -4686027666808657977 Time: 0.595131 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.565781 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.590729 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.575729 [TRT] Fastest Tactic: 4772821744921268633 Time: 0.458046 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 4772821744921268633 [TRT] *************** Autotuning Reformat:Float(200704,784,28,1) -> Float(200704,1,7168,256) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.122136 [TRT] Tactic: 0 Time: 0.178698 [TRT] Fastest Tactic: 1002 Time: 0.122136 [TRT] *************** Autotuning Reformat:Float(200704,784,28,1) -> Half(200704,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.214375 [TRT] Tactic: 0 Time: 0.155468 [TRT] Fastest Tactic: 0 Time: 0.155468 [TRT] *************** Autotuning Reformat:Float(200704,784,28,1) -> Half(100352,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.166589 [TRT] Tactic: 0 Time: 0.0948695 [TRT] Fastest Tactic: 0 Time: 0.0948695 [TRT] *************** Autotuning Reformat:Float(200704,1,7168,256) -> Float(200704,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.150677 [TRT] Tactic: 0 Time: 0.173203 [TRT] Fastest Tactic: 1002 Time: 0.150677 [TRT] *************** Autotuning Reformat:Float(200704,1,7168,256) -> Half(200704,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.111823 [TRT] Tactic: 0 Time: 0.171145 [TRT] Fastest Tactic: 1002 Time: 0.111823 [TRT] *************** Autotuning Reformat:Float(200704,1,7168,256) -> Half(100352,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.147005 [TRT] Tactic: 0 Time: 0.194115 [TRT] Fastest Tactic: 1002 Time: 0.147005 [TRT] *************** Autotuning Reformat:Half(200704,784,28,1) -> Float(200704,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.216537 [TRT] Tactic: 0 Time: 0.155677 [TRT] Fastest Tactic: 0 Time: 0.155677 [TRT] *************** Autotuning Reformat:Half(200704,784,28,1) -> Float(200704,1,7168,256) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.105312 [TRT] Tactic: 0 Time: 0.170833 [TRT] Fastest Tactic: 1002 Time: 0.105312 [TRT] *************** Autotuning Reformat:Half(200704,784,28,1) -> Half(100352,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.113177 [TRT] Tactic: 0 Time: 0.093958 [TRT] Fastest Tactic: 0 Time: 0.093958 [TRT] *************** Autotuning Reformat:Half(100352,784:2,28,1) -> Float(200704,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.139636 [TRT] Tactic: 0 Time: 0.182865 [TRT] Fastest Tactic: 1002 Time: 0.139636 [TRT] *************** Autotuning Reformat:Half(100352,784:2,28,1) -> Float(200704,1,7168,256) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.10586 [TRT] Tactic: 0 Time: 0.196589 [TRT] Fastest Tactic: 1002 Time: 0.10586 [TRT] *************** Autotuning Reformat:Half(100352,784:2,28,1) -> Half(200704,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.197266 [TRT] Tactic: 0 Time: 0.164973 [TRT] Fastest Tactic: 0 Time: 0.164973 [TRT] *************** Autotuning Reformat:Float(137984,784,28,1) -> Float(137984,1,4928,176) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0652085 [TRT] Tactic: 0 Time: 0.023854 [TRT] Fastest Tactic: 0 Time: 0.023854 [TRT] *************** Autotuning Reformat:Float(137984,784,28,1) -> Half(137984,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.031458 [TRT] Tactic: 0 Time: 0.0230735 [TRT] Fastest Tactic: 0 Time: 0.0230735 [TRT] *************** Autotuning Reformat:Float(137984,784,28,1) -> Half(68992,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0338025 [TRT] Tactic: 0 Time: 0.026094 [TRT] Fastest Tactic: 0 Time: 0.026094 [TRT] *************** Autotuning Reformat:Float(137984,1,4928,176) -> Float(137984,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0652345 [TRT] Tactic: 0 Time: 0.0237755 [TRT] Fastest Tactic: 0 Time: 0.0237755 [TRT] *************** Autotuning Reformat:Float(137984,1,4928,176) -> Half(137984,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0333595 [TRT] Tactic: 0 Time: 0.0238545 [TRT] Fastest Tactic: 0 Time: 0.0238545 [TRT] *************** Autotuning Reformat:Float(137984,1,4928,176) -> Half(68992,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.035052 [TRT] Tactic: 0 Time: 0.0264325 [TRT] Fastest Tactic: 0 Time: 0.0264325 [TRT] *************** Autotuning Reformat:Half(137984,784,28,1) -> Float(137984,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.03151 [TRT] Tactic: 0 Time: 0.0158855 [TRT] Fastest Tactic: 0 Time: 0.0158855 [TRT] *************** Autotuning Reformat:Half(137984,784,28,1) -> Float(137984,1,4928,176) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0328125 [TRT] Tactic: 0 Time: 0.0239845 [TRT] Fastest Tactic: 0 Time: 0.0239845 [TRT] *************** Autotuning Reformat:Half(137984,784,28,1) -> Half(68992,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0184635 [TRT] Tactic: 0 Time: 0.015703 [TRT] Fastest Tactic: 0 Time: 0.015703 [TRT] *************** Autotuning Reformat:Half(68992,784:2,28,1) -> Float(137984,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0337235 [TRT] Tactic: 0 Time: 0.013307 [TRT] Fastest Tactic: 0 Time: 0.013307 [TRT] *************** Autotuning Reformat:Half(68992,784:2,28,1) -> Float(137984,1,4928,176) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0333855 [TRT] Tactic: 0 Time: 0.028203 [TRT] Fastest Tactic: 0 Time: 0.028203 [TRT] *************** Autotuning Reformat:Half(68992,784:2,28,1) -> Half(137984,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0781255 [TRT] Tactic: 0 Time: 0.023672 [TRT] Fastest Tactic: 0 Time: 0.023672 [TRT] *************** Autotuning format combination: Float(137984,784,28,1) -> Float(200704,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/5x5 + inception_3a/relu_5x5 (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3a/5x5 + inception_3a/relu_5x5 (FusedConvActConvolution) [TRT] Tactic: 393215 Time: 0.171927 [TRT] Tactic: 917503 Time: 0.185859 [TRT] Tactic: 1114111 Time: 0.287422 [TRT] Tactic: 1245183 Time: 0.293463 [TRT] Tactic: 1572863 Time: 0.180468 [TRT] Tactic: 2490367 Time: 0.184766 [TRT] Tactic: 2555903 Time: 0.194348 [TRT] Tactic: 2949119 Time: 0.308334 [TRT] Tactic: 3211263 Time: 0.599193 [TRT] Tactic: 3801087 Time: 0.201641 [TRT] Tactic: 3866623 Time: 0.180442 [TRT] Tactic: 4128767 Time: 0.296953 [TRT] Tactic: 4456447 Time: 0.181406 [TRT] Tactic: 4718591 Time: 0.276484 [TRT] Tactic: 4784127 Time: 0.584922 [TRT] Tactic: 4849663 Time: 0.270833 [TRT] Tactic: 5111807 Time: 0.268907 [TRT] Tactic: 5308415 Time: 0.292448 [TRT] Tactic: 5505023 Time: 0.568229 [TRT] Tactic: 6094847 Time: 0.212005 [TRT] Tactic: 6356991 Time: 0.190026 [TRT] Tactic: 6553599 Time: 0.173177 [TRT] Tactic: 6619135 Time: 0.208229 [TRT] Tactic: 6684671 Time: 0.54112 [TRT] Tactic: 7471103 Time: 0.184505 [TRT] Tactic: 7667711 Time: 0.276927 [TRT] Tactic: 7929855 Time: 0.286171 [TRT] Tactic: 8060927 Time: 0.185026 [TRT] Tactic: 8126463 Time: 0.32375 [TRT] Tactic: 8388607 Time: 0.305026 [TRT] Tactic: 8519679 Time: 0.221927 [TRT] Tactic: 8781823 Time: 0.361042 [TRT] Tactic: 8912895 Time: 0.348802 [TRT] Tactic: 9240575 Time: 0.289401 [TRT] Tactic: 9306111 Time: 0.273463 [TRT] Tactic: 9371647 Time: 0.272317 [TRT] Tactic: 9437183 Time: 0.309115 [TRT] Tactic: 9633791 Time: 0.270286 [TRT] Tactic: 9699327 Time: 0.177109 [TRT] Tactic: 9764863 Time: 0.169479 [TRT] Tactic: 10158079 Time: 0.187917 [TRT] Tactic: 10420223 Time: 0.30737 [TRT] Tactic: 10616831 Time: 0.202291 [TRT] Tactic: 10878975 Time: 0.182708 [TRT] Fastest Tactic: 9764863 Time: 0.169479 [TRT] --------------- Timing Runner: inception_3a/5x5 + inception_3a/relu_5x5 (CudnnConvolution) [TRT] Tactic: 0 Time: 0.344297 [TRT] Tactic: 1 Time: 0.346146 [TRT] Tactic: 2 Time: 0.863177 [TRT] Tactic: 4 Time: 1.19424 [TRT] Tactic: 5 Time: 1.20669 [TRT] Fastest Tactic: 0 Time: 0.344297 [TRT] --------------- Timing Runner: inception_3a/5x5 + inception_3a/relu_5x5 (CaskConvolution) [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.233958 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v0 Tactic: 1754984623894446479 [TRT] Tactic: 1754984623894446479 Time: 0.23026 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v0 Tactic: 3611739942397549984 [TRT] Tactic: 3611739942397549984 Time: 0.573151 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v1 Tactic: 4337000649858996379 [TRT] Tactic: 4337000649858996379 Time: 0.307449 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.567343 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.278255 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.54573 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.302916 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.179193 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v1 Tactic: -9137461792520977713 [TRT] Tactic: -9137461792520977713 Time: 0.58112 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.574323 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v0 Tactic: -8133971918129952780 [TRT] Tactic: -8133971918129952780 Time: 0.304323 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v1 Tactic: -6092040395344634144 [TRT] Tactic: -6092040395344634144 Time: 0.245286 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.22638 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.193541 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.302552 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.288958 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.534505 [TRT] Fastest Tactic: 7144526460361122478 Time: 0.179193 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 9764863 [TRT] *************** Autotuning format combination: Float(137984,1,4928,176) -> Float(200704,1,7168,256) *************** [TRT] --------------- Timing Runner: inception_3a/5x5 + inception_3a/relu_5x5 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3a/5x5 + inception_3a/relu_5x5 (CaskConvolution) [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.349818 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.273072 [TRT] Fastest Tactic: -7394439838318485025 Time: 0.273072 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(137984,784,28,1) -> Half(200704,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/5x5 + inception_3a/relu_5x5 (CudnnConvolution) [TRT] Tactic: 0 Time: 0.352161 [TRT] Tactic: 1 Time: 0.351433 [TRT] Tactic: 2 Time: 0.852995 [TRT] Tactic: 4 Time: 1.18815 [TRT] Tactic: 5 Time: 1.20279 [TRT] Fastest Tactic: 1 Time: 0.351433 [TRT] --------------- Timing Runner: inception_3a/5x5 + inception_3a/relu_5x5 (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 1 [TRT] *************** Autotuning format combination: Half(68992,784:2,28,1) -> Half(100352,784:2,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/5x5 + inception_3a/relu_5x5 (FusedConvActConvolution) [TRT] Tactic: 393215 Time: 0.096068 [TRT] Tactic: 917503 Time: 0.112214 [TRT] Tactic: 1114111 Time: 0.162578 [TRT] Tactic: 1245183 Time: 0.169427 [TRT] Tactic: 1572863 Time: 0.10474 [TRT] Tactic: 2490367 Time: 0.124583 [TRT] Tactic: 2555903 Time: 0.136615 [TRT] Tactic: 2949119 Time: 0.212813 [TRT] Tactic: 3211263 Time: 0.35026 [TRT] Tactic: 3801087 Time: 0.112839 [TRT] Tactic: 3866623 Time: 0.102839 [TRT] Tactic: 4128767 Time: 0.186693 [TRT] Tactic: 4456447 Time: 0.110546 [TRT] Tactic: 4718591 Time: 0.160417 [TRT] Tactic: 4784127 Time: 0.341173 [TRT] Tactic: 4849663 Time: 0.152552 [TRT] Tactic: 5111807 Time: 0.153151 [TRT] Tactic: 5308415 Time: 0.1725 [TRT] Tactic: 5505023 Time: 0.363099 [TRT] Tactic: 6094847 Time: 0.118803 [TRT] Tactic: 6356991 Time: 0.134635 [TRT] Tactic: 6553599 Time: 0.0951305 [TRT] Tactic: 6619135 Time: 0.106511 [TRT] Tactic: 6684671 Time: 0.340052 [TRT] Tactic: 7471103 Time: 0.107838 [TRT] Tactic: 7667711 Time: 0.160782 [TRT] Tactic: 7929855 Time: 0.159375 [TRT] Tactic: 8060927 Time: 0.106901 [TRT] Tactic: 8126463 Time: 0.175391 [TRT] Tactic: 8388607 Time: 0.186563 [TRT] Tactic: 8519679 Time: 0.122473 [TRT] Tactic: 8781823 Time: 0.189609 [TRT] Tactic: 8912895 Time: 0.217552 [TRT] Tactic: 9240575 Time: 0.167188 [TRT] Tactic: 9306111 Time: 0.137735 [TRT] Tactic: 9371647 Time: 0.15323 [TRT] Tactic: 9437183 Time: 0.213203 [TRT] Tactic: 9633791 Time: 0.152448 [TRT] Tactic: 9699327 Time: 0.0989845 [TRT] Tactic: 9764863 Time: 0.100859 [TRT] Tactic: 10158079 Time: 0.104219 [TRT] Tactic: 10420223 Time: 0.217604 [TRT] Tactic: 10616831 Time: 0.113568 [TRT] Tactic: 10878975 Time: 0.106718 [TRT] Fastest Tactic: 6553599 Time: 0.0951305 [TRT] --------------- Timing Runner: inception_3a/5x5 + inception_3a/relu_5x5 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3a/5x5 + inception_3a/relu_5x5 (CaskConvolution) [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.126588 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_large_nn_v1 Tactic: 3650389455493082349 [TRT] Tactic: 3650389455493082349 Time: 0.131979 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.106042 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.163437 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_large_nn_v1 Tactic: -6490690591794140522 [TRT] Tactic: -6490690591794140522 Time: 0.164166 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_large_nn_v1 Tactic: -4686027666808657977 [TRT] Tactic: -4686027666808657977 Time: 0.299818 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.275156 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.29487 [TRT] inception_3a/5x5 + inception_3a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.147578 [TRT] Fastest Tactic: 5319956359050645452 Time: 0.106042 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 6553599 [TRT] *************** Autotuning Reformat:Float(200704,784,28,1) -> Float(200704,1,7168,256) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.06526 [TRT] Tactic: 0 Time: 0.044713 [TRT] Fastest Tactic: 0 Time: 0.044713 [TRT] *************** Autotuning Reformat:Float(200704,784,28,1) -> Half(200704,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.057526 [TRT] Tactic: 0 Time: 0.041641 [TRT] Fastest Tactic: 0 Time: 0.041641 [TRT] *************** Autotuning Reformat:Float(200704,784,28,1) -> Half(100352,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.046953 [TRT] Tactic: 0 Time: 0.026016 [TRT] Fastest Tactic: 0 Time: 0.026016 [TRT] *************** Autotuning Reformat:Float(200704,1,7168,256) -> Float(200704,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0676565 [TRT] Tactic: 0 Time: 0.041979 [TRT] Fastest Tactic: 0 Time: 0.041979 [TRT] *************** Autotuning Reformat:Float(200704,1,7168,256) -> Half(200704,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.031458 [TRT] Tactic: 0 Time: 0.0421355 [TRT] Fastest Tactic: 1002 Time: 0.031458 [TRT] *************** Autotuning Reformat:Float(200704,1,7168,256) -> Half(100352,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.043672 [TRT] Tactic: 0 Time: 0.049557 [TRT] Fastest Tactic: 1002 Time: 0.043672 [TRT] *************** Autotuning Reformat:Half(200704,784,28,1) -> Float(200704,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.059141 [TRT] Tactic: 0 Time: 0.0419795 [TRT] Fastest Tactic: 0 Time: 0.0419795 [TRT] *************** Autotuning Reformat:Half(200704,784,28,1) -> Float(200704,1,7168,256) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0303645 [TRT] Tactic: 0 Time: 0.046927 [TRT] Fastest Tactic: 1002 Time: 0.0303645 [TRT] *************** Autotuning Reformat:Half(200704,784,28,1) -> Half(100352,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.03125 [TRT] Tactic: 0 Time: 0.0261715 [TRT] Fastest Tactic: 0 Time: 0.0261715 [TRT] *************** Autotuning Reformat:Half(100352,784:2,28,1) -> Float(200704,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0414585 [TRT] Tactic: 0 Time: 0.0487245 [TRT] Fastest Tactic: 1002 Time: 0.0414585 [TRT] *************** Autotuning Reformat:Half(100352,784:2,28,1) -> Float(200704,1,7168,256) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.031068 [TRT] Tactic: 0 Time: 0.05237 [TRT] Fastest Tactic: 1002 Time: 0.031068 [TRT] *************** Autotuning Reformat:Half(100352,784:2,28,1) -> Half(200704,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.081589 [TRT] Tactic: 0 Time: 0.044453 [TRT] Fastest Tactic: 0 Time: 0.044453 [TRT] *************** Autotuning Reformat:Float(150528,784,28,1) -> Half(150528,784,28,1) *************** [TRT] *************** Autotuning Reformat:Float(150528,784,28,1) -> Half(75264,784:2,28,1) *************** [TRT] *************** Autotuning Reformat:Float(150528,1,5376,192) -> Float(150528,784,28,1) *************** [TRT] *************** Autotuning Reformat:Float(150528,1,5376,192) -> Half(150528,784,28,1) *************** [TRT] *************** Autotuning Reformat:Float(150528,1,5376,192) -> Half(75264,784:2,28,1) *************** [TRT] *************** Autotuning Reformat:Half(150528,784,28,1) -> Float(150528,784,28,1) *************** [TRT] *************** Autotuning Reformat:Half(150528,784,28,1) -> Half(75264,784:2,28,1) *************** [TRT] *************** Autotuning Reformat:Half(75264,784:2,28,1) -> Float(150528,784,28,1) *************** [TRT] *************** Autotuning Reformat:Half(75264,784:2,28,1) -> Half(150528,784,28,1) *************** [TRT] *************** Autotuning format combination: Float(150528,784,28,1) -> Float(150528,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/pool (TiledPooling) [TRT] Tactic: 2752769 Time: 0.352292 [TRT] Tactic: 2818305 Time: 0.331901 [TRT] Tactic: 2883841 Time: 0.250312 [TRT] Tactic: 2949377 Time: 0.692605 [TRT] Tactic: 3014913 Time: 0.548542 [TRT] Tactic: 3080449 Time: 0.299245 [TRT] Tactic: 3145985 Time: 0.276146 [TRT] Tactic: 3211521 Time: 0.246459 [TRT] Tactic: 3277057 Time: 0.241406 [TRT] Tactic: 3342593 Time: 0.17586 [TRT] Tactic: 3408129 Time: 0.41487 [TRT] Tactic: 3473665 Time: 0.33487 [TRT] Tactic: 3539201 Time: 0.189974 [TRT] Tactic: 3604737 Time: 0.182214 [TRT] Tactic: 3670273 Time: 0.248151 [TRT] Tactic: 3735809 Time: 0.237604 [TRT] Tactic: 3801345 Time: 0.152474 [TRT] Tactic: 3866881 Time: 0.334063 [TRT] Tactic: 3932417 Time: 0.267291 [TRT] Tactic: 3997953 Time: 0.157448 [TRT] Tactic: 4063489 Time: 0.151198 [TRT] Tactic: 4129025 Time: 0.250078 [TRT] Tactic: 4194561 Time: 0.23901 [TRT] Tactic: 4260097 Time: 0.150338 [TRT] Tactic: 4325633 Time: 0.289974 [TRT] Tactic: 4391169 Time: 0.233308 [TRT] Tactic: 4456705 Time: 0.139297 [TRT] Tactic: 4522241 Time: 0.139766 [TRT] Tactic: 4587777 Time: 0.252422 [TRT] Tactic: 4653313 Time: 0.240339 [TRT] Tactic: 4718849 Time: 0.150079 [TRT] Tactic: 4784385 Time: 0.271562 [TRT] Tactic: 4849921 Time: 0.223516 [TRT] Tactic: 4915457 Time: 0.132787 [TRT] Tactic: 4980993 Time: 0.134036 [TRT] Tactic: 5046529 Time: 0.249713 [TRT] Tactic: 5112065 Time: 0.241719 [TRT] Tactic: 5177601 Time: 0.150469 [TRT] Tactic: 5243137 Time: 0.253125 [TRT] Tactic: 5308673 Time: 0.207005 [TRT] Tactic: 5374209 Time: 0.12789 [TRT] Tactic: 5439745 Time: 0.130989 [TRT] Tactic: 6553857 Time: 0.121485 [TRT] Tactic: 6750465 Time: 0.178959 [TRT] Fastest Tactic: 6553857 Time: 0.121485 [TRT] --------------- Timing Runner: inception_3a/pool (CudnnPooling) [TRT] Tactic: -1 Time: 0.497474 [TRT] Fastest Tactic: -1 Time: 0.497474 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: TiledPooling Tactic: 6553857 [TRT] *************** Autotuning format combination: Half(150528,784,28,1) -> Half(150528,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/pool (TiledPooling) [TRT] TiledPooling has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3a/pool (CudnnPooling) [TRT] Tactic: -1 Time: 0.519167 [TRT] Fastest Tactic: -1 Time: 0.519167 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnPooling Tactic: -1 [TRT] *************** Autotuning format combination: Half(75264,784:2,28,1) -> Half(75264,784:2,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/pool (TiledPooling) [TRT] Tactic: 2752769 Time: 0.196198 [TRT] Tactic: 2818305 Time: 0.192943 [TRT] Tactic: 2883841 Time: 0.14724 [TRT] Tactic: 2949377 Time: 0.380704 [TRT] Tactic: 3014913 Time: 0.315183 [TRT] Tactic: 3080449 Time: 0.174713 [TRT] Tactic: 3145985 Time: 0.157787 [TRT] Tactic: 3211521 Time: 0.13875 [TRT] Tactic: 3277057 Time: 0.134818 [TRT] Tactic: 3342593 Time: 0.10987 [TRT] Tactic: 3408129 Time: 0.257969 [TRT] Tactic: 3473665 Time: 0.210781 [TRT] Tactic: 3539201 Time: 0.118672 [TRT] Tactic: 3604737 Time: 0.113542 [TRT] Tactic: 3670273 Time: 0.122735 [TRT] Tactic: 3735809 Time: 0.118255 [TRT] Tactic: 3801345 Time: 0.094948 [TRT] Tactic: 3866881 Time: 0.223516 [TRT] Tactic: 3932417 Time: 0.181641 [TRT] Tactic: 3997953 Time: 0.103672 [TRT] Tactic: 4063489 Time: 0.099141 [TRT] Tactic: 4129025 Time: 0.118307 [TRT] Tactic: 4194561 Time: 0.114063 [TRT] Tactic: 4260097 Time: 0.0891665 [TRT] Tactic: 4325633 Time: 0.199323 [TRT] Tactic: 4391169 Time: 0.167161 [TRT] Tactic: 4456705 Time: 0.0975515 [TRT] Tactic: 4522241 Time: 0.0917705 [TRT] Tactic: 4587777 Time: 0.119557 [TRT] Tactic: 4653313 Time: 0.115208 [TRT] Tactic: 4718849 Time: 0.08987 [TRT] Tactic: 4784385 Time: 0.198307 [TRT] Tactic: 4849921 Time: 0.164245 [TRT] Tactic: 4915457 Time: 0.0947135 [TRT] Tactic: 4980993 Time: 0.0924215 [TRT] Tactic: 5046529 Time: 0.119922 [TRT] Tactic: 5112065 Time: 0.114688 [TRT] Tactic: 5177601 Time: 0.0859115 [TRT] Tactic: 5243137 Time: 0.185364 [TRT] Tactic: 5308673 Time: 0.152813 [TRT] Tactic: 5374209 Time: 0.0891405 [TRT] Tactic: 5439745 Time: 0.086823 [TRT] Tactic: 6553857 Time: 0.078594 [TRT] Tactic: 6750465 Time: 0.119505 [TRT] Fastest Tactic: 6553857 Time: 0.078594 [TRT] --------------- Timing Runner: inception_3a/pool (CudaPooling) [TRT] Tactic: -3 Time: 0.29302 [TRT] Fastest Tactic: -3 Time: 0.29302 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: TiledPooling Tactic: 6553857 [TRT] *************** Autotuning Reformat:Float(150528,784,28,1) -> Float(150528,1,5376,192) *************** [TRT] *************** Autotuning Reformat:Float(150528,784,28,1) -> Half(150528,784,28,1) *************** [TRT] *************** Autotuning Reformat:Float(150528,784,28,1) -> Half(75264,784:2,28,1) *************** [TRT] *************** Autotuning Reformat:Float(150528,1,5376,192) -> Float(150528,784,28,1) *************** [TRT] *************** Autotuning Reformat:Float(150528,1,5376,192) -> Half(150528,784,28,1) *************** [TRT] *************** Autotuning Reformat:Float(150528,1,5376,192) -> Half(75264,784:2,28,1) *************** [TRT] *************** Autotuning Reformat:Half(150528,784,28,1) -> Float(150528,784,28,1) *************** [TRT] *************** Autotuning Reformat:Half(150528,784,28,1) -> Float(150528,1,5376,192) *************** [TRT] *************** Autotuning Reformat:Half(150528,784,28,1) -> Half(75264,784:2,28,1) *************** [TRT] *************** Autotuning Reformat:Half(75264,784:2,28,1) -> Float(150528,784,28,1) *************** [TRT] *************** Autotuning Reformat:Half(75264,784:2,28,1) -> Float(150528,1,5376,192) *************** [TRT] *************** Autotuning Reformat:Half(75264,784:2,28,1) -> Half(150528,784,28,1) *************** [TRT] *************** Autotuning format combination: Float(150528,784,28,1) -> Float(200704,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/pool_proj + inception_3a/relu_pool_proj (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3a/pool_proj + inception_3a/relu_pool_proj (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.175624 [TRT] Tactic: 655359 Time: 0.377422 [TRT] Tactic: 786431 Time: 0.212084 [TRT] Tactic: 851967 Time: 0.506172 [TRT] Tactic: 1179647 Time: 0.29086 [TRT] Tactic: 1310719 Time: 0.344974 [TRT] Tactic: 1376255 Time: 0.245703 [TRT] Tactic: 1441791 Time: 0.315547 [TRT] Tactic: 1507327 Time: 0.496615 [TRT] Tactic: 1638399 Time: 0.215052 [TRT] Tactic: 1835007 Time: 0.197682 [TRT] Tactic: 1900543 Time: 0.495234 [TRT] Tactic: 2097151 Time: 0.212032 [TRT] Tactic: 2162687 Time: 0.248073 [TRT] Tactic: 2293759 Time: 0.255859 [TRT] Tactic: 2359295 Time: 0.258723 [TRT] Tactic: 2686975 Time: 0.233281 [TRT] Tactic: 3080191 Time: 0.399609 [TRT] Tactic: 3342335 Time: 0.498203 [TRT] Tactic: 3407871 Time: 0.259844 [TRT] Tactic: 3538943 Time: 0.257006 [TRT] Tactic: 3670015 Time: 0.345573 [TRT] Tactic: 3932159 Time: 0.448047 [TRT] Tactic: 3997695 Time: 0.214635 [TRT] Tactic: 4063231 Time: 0.496354 [TRT] Tactic: 4194303 Time: 0.186379 [TRT] Tactic: 4259839 Time: 0.220886 [TRT] Tactic: 4325375 Time: 0.20039 [TRT] Tactic: 4521983 Time: 0.246249 [TRT] Tactic: 4587519 Time: 0.224791 [TRT] Tactic: 4653055 Time: 0.357656 [TRT] Tactic: 4915199 Time: 0.186511 [TRT] Tactic: 4980735 Time: 0.2025 [TRT] Tactic: 5177343 Time: 0.282057 [TRT] Tactic: 5242879 Time: 0.234271 [TRT] Tactic: 5373951 Time: 0.397578 [TRT] Tactic: 5439487 Time: 0.275052 [TRT] Tactic: 5570559 Time: 0.336329 [TRT] Tactic: 5636095 Time: 0.498776 [TRT] Tactic: 5701631 Time: 0.27474 [TRT] Tactic: 5767167 Time: 0.476849 [TRT] Tactic: 5832703 Time: 0.254323 [TRT] Tactic: 5898239 Time: 0.22276 [TRT] Tactic: 6029311 Time: 0.239792 [TRT] Tactic: 6225919 Time: 0.228047 [TRT] Tactic: 6291455 Time: 0.291328 [TRT] Tactic: 6422527 Time: 0.404218 [TRT] Tactic: 6750207 Time: 0.197891 [TRT] Tactic: 6815743 Time: 0.289349 [TRT] Tactic: 6946815 Time: 0.282135 [TRT] Tactic: 7012351 Time: 0.213281 [TRT] Tactic: 7077887 Time: 0.251952 [TRT] Tactic: 7143423 Time: 0.300547 [TRT] Tactic: 7208959 Time: 0.24724 [TRT] Tactic: 7340031 Time: 0.234505 [TRT] Tactic: 7405567 Time: 0.278958 [TRT] Tactic: 7536639 Time: 0.284297 [TRT] Tactic: 7602175 Time: 0.238854 [TRT] Tactic: 7733247 Time: 0.247213 [TRT] Tactic: 7798783 Time: 0.212057 [TRT] Tactic: 8191999 Time: 0.284037 [TRT] Tactic: 8257535 Time: 0.192162 [TRT] Tactic: 8323071 Time: 0.205156 [TRT] Tactic: 8650751 Time: 0.246875 [TRT] Tactic: 8716287 Time: 0.301692 [TRT] Tactic: 9109503 Time: 0.223698 [TRT] Tactic: 9568255 Time: 0.185235 [TRT] Tactic: 9895935 Time: 0.184192 [TRT] Tactic: 10223615 Time: 0.230365 [TRT] Tactic: 10354687 Time: 0.209662 [TRT] Tactic: 10551295 Time: 0.185 [TRT] Tactic: 10747903 Time: 0.233776 [TRT] Tactic: 10944511 Time: 0.203021 [TRT] Fastest Tactic: 589823 Time: 0.175624 [TRT] --------------- Timing Runner: inception_3a/pool_proj + inception_3a/relu_pool_proj (CudnnConvolution) [TRT] Tactic: 0 Time: 0.192318 [TRT] Tactic: 1 Time: 0.193464 [TRT] Tactic: 2 Time: 0.558984 [TRT] Tactic: 4 skipped. Scratch requested: 55173120, available: 33554432 [TRT] Tactic: 5 Time: 0.776667 [TRT] Fastest Tactic: 0 Time: 0.192318 [TRT] --------------- Timing Runner: inception_3a/pool_proj + inception_3a/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3a/pool_proj + inception_3a/relu_pool_proj (CaskConvolution) [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.124479 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v0 Tactic: 1698681053543049347 [TRT] Tactic: 1698681053543049347 Time: 0.112032 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.314792 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.162006 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.315834 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v1 Tactic: 5326823351883942011 [TRT] Tactic: 5326823351883942011 Time: 0.302318 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v0 Tactic: 5500448035057547314 [TRT] Tactic: 5500448035057547314 Time: 0.178307 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.170208 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.126328 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.316172 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v1 Tactic: -6576203419454146580 [TRT] Tactic: -6576203419454146580 Time: 0.119453 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.132708 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.125234 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.181536 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.175547 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.305755 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v0 Tactic: -377491875521947884 [TRT] Tactic: -377491875521947884 Time: 0.31185 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v1 Tactic: -37215280111360163 [TRT] Tactic: -37215280111360163 Time: 0.162083 [TRT] Fastest Tactic: 1698681053543049347 Time: 0.112032 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 1698681053543049347 [TRT] *************** Autotuning format combination: Float(150528,1,5376,192) -> Float(200704,1,7168,256) *************** [TRT] --------------- Timing Runner: inception_3a/pool_proj + inception_3a/relu_pool_proj (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3a/pool_proj + inception_3a/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3a/pool_proj + inception_3a/relu_pool_proj (CaskConvolution) [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 3886731678879822788 [TRT] Tactic: 3886731678879822788 Time: 0.167084 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 6629944304117643200 [TRT] Tactic: 6629944304117643200 Time: 0.163099 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.16513 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.17138 [TRT] Fastest Tactic: 6629944304117643200 Time: 0.163099 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 6629944304117643200 [TRT] *************** Autotuning format combination: Half(150528,784,28,1) -> Half(200704,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/pool_proj + inception_3a/relu_pool_proj (CudnnConvolution) [TRT] Tactic: 0 Time: 0.212083 [TRT] Tactic: 1 Time: 0.195885 [TRT] Tactic: 2 Time: 0.591978 [TRT] Tactic: 4 skipped. Scratch requested: 55173120, available: 33554432 [TRT] Tactic: 5 Time: 0.776771 [TRT] Fastest Tactic: 1 Time: 0.195885 [TRT] --------------- Timing Runner: inception_3a/pool_proj + inception_3a/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3a/pool_proj + inception_3a/relu_pool_proj (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 1 [TRT] *************** Autotuning format combination: Half(75264,784:2,28,1) -> Half(200704,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/pool_proj + inception_3a/relu_pool_proj (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] *************** Autotuning format combination: Half(75264,784:2,28,1) -> Half(100352,784:2,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/pool_proj + inception_3a/relu_pool_proj (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.120078 [TRT] Tactic: 655359 Time: 0.28414 [TRT] Tactic: 786431 Time: 0.148724 [TRT] Tactic: 851967 Time: 0.343776 [TRT] Tactic: 1179647 Time: 0.184375 [TRT] Tactic: 1310719 Time: 0.194193 [TRT] Tactic: 1376255 Time: 0.13987 [TRT] Tactic: 1441791 Time: 0.208046 [TRT] Tactic: 1507327 Time: 0.316406 [TRT] Tactic: 1638399 Time: 0.130652 [TRT] Tactic: 1835007 Time: 0.139739 [TRT] Tactic: 1900543 Time: 0.2869 [TRT] Tactic: 2097151 Time: 0.150313 [TRT] Tactic: 2162687 Time: 0.168855 [TRT] Tactic: 2293759 Time: 0.177344 [TRT] Tactic: 2359295 Time: 0.172422 [TRT] Tactic: 2686975 Time: 0.24198 [TRT] Tactic: 3080191 Time: 0.319036 [TRT] Tactic: 3342335 Time: 0.347266 [TRT] Tactic: 3407871 Time: 0.179583 [TRT] Tactic: 3538943 Time: 0.189349 [TRT] Tactic: 3670015 Time: 0.315625 [TRT] Tactic: 3932159 Time: 0.259557 [TRT] Tactic: 3997695 Time: 0.159348 [TRT] Tactic: 4063231 Time: 0.375833 [TRT] Tactic: 4194303 Time: 0.134245 [TRT] Tactic: 4259839 Time: 0.162031 [TRT] Tactic: 4325375 Time: 0.131719 [TRT] Tactic: 4521983 Time: 0.148959 [TRT] Tactic: 4587519 Time: 0.168072 [TRT] Tactic: 4653055 Time: 0.268698 [TRT] Tactic: 4915199 Time: 0.129791 [TRT] Tactic: 4980735 Time: 0.144166 [TRT] Tactic: 5177343 Time: 0.179505 [TRT] Tactic: 5242879 Time: 0.152969 [TRT] Tactic: 5373951 Time: 0.21336 [TRT] Tactic: 5439487 Time: 0.165052 [TRT] Tactic: 5570559 Time: 0.32539 [TRT] Tactic: 5636095 Time: 0.374219 [TRT] Tactic: 5701631 Time: 0.150286 [TRT] Tactic: 5767167 Time: 0.267578 [TRT] Tactic: 5832703 Time: 0.167291 [TRT] Tactic: 5898239 Time: 0.200026 [TRT] Tactic: 6029311 Time: 0.162473 [TRT] Tactic: 6225919 Time: 0.15685 [TRT] Tactic: 6291455 Time: 0.202656 [TRT] Tactic: 6422527 Time: 0.289193 [TRT] Tactic: 6750207 Time: 0.137656 [TRT] Tactic: 6815743 Time: 0.182526 [TRT] Tactic: 6946815 Time: 0.173984 [TRT] Tactic: 7012351 Time: 0.183542 [TRT] Tactic: 7077887 Time: 0.186536 [TRT] Tactic: 7143423 Time: 0.203855 [TRT] Tactic: 7208959 Time: 0.173099 [TRT] Tactic: 7340031 Time: 0.222578 [TRT] Tactic: 7405567 Time: 0.214036 [TRT] Tactic: 7536639 Time: 0.206876 [TRT] Tactic: 7602175 Time: 0.156146 [TRT] Tactic: 7733247 Time: 0.189557 [TRT] Tactic: 7798783 Time: 0.179947 [TRT] Tactic: 8191999 Time: 0.198177 [TRT] Tactic: 8257535 Time: 0.141614 [TRT] Tactic: 8323071 Time: 0.147265 [TRT] Tactic: 8650751 Time: 0.15875 [TRT] Tactic: 8716287 Time: 0.199739 [TRT] Tactic: 9109503 Time: 0.176589 [TRT] Tactic: 9568255 Time: 0.143698 [TRT] Tactic: 9895935 Time: 0.148307 [TRT] Tactic: 10223615 Time: 0.263177 [TRT] Tactic: 10354687 Time: 0.175833 [TRT] Tactic: 10551295 Time: 0.127344 [TRT] Tactic: 10747903 Time: 0.182005 [TRT] Tactic: 10944511 Time: 0.159193 [TRT] Fastest Tactic: 589823 Time: 0.120078 [TRT] --------------- Timing Runner: inception_3a/pool_proj + inception_3a/relu_pool_proj (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3a/pool_proj + inception_3a/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3a/pool_proj + inception_3a/relu_pool_proj (CaskConvolution) [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_interior_nn_v1 Tactic: 3066127711859985668 [TRT] Tactic: 3066127711859985668 Time: 0.080286 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.093542 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.085391 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.117579 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] Tactic: 8163473458334948789 Time: 0.113385 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.202969 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.208854 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.113646 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_interior_nn_v1 Tactic: -1716393687483585322 [TRT] Tactic: -1716393687483585322 Time: 0.200599 [TRT] Fastest Tactic: 3066127711859985668 Time: 0.080286 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 3066127711859985668 [TRT] *************** Autotuning Reformat:Float(200704,784,28,1) -> Float(200704,1,7168,256) *************** [TRT] *************** Autotuning Reformat:Float(200704,784,28,1) -> Half(200704,784,28,1) *************** [TRT] *************** Autotuning Reformat:Float(200704,784,28,1) -> Half(100352,784:2,28,1) *************** [TRT] *************** Autotuning Reformat:Float(200704,1,7168,256) -> Float(200704,784,28,1) *************** [TRT] *************** Autotuning Reformat:Float(200704,1,7168,256) -> Half(200704,784,28,1) *************** [TRT] *************** Autotuning Reformat:Float(200704,1,7168,256) -> Half(100352,784:2,28,1) *************** [TRT] *************** Autotuning Reformat:Half(200704,784,28,1) -> Float(200704,784,28,1) *************** [TRT] *************** Autotuning Reformat:Half(200704,784,28,1) -> Float(200704,1,7168,256) *************** [TRT] *************** Autotuning Reformat:Half(200704,784,28,1) -> Half(100352,784:2,28,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,784:2,28,1) -> Float(200704,784,28,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,784:2,28,1) -> Float(200704,1,7168,256) *************** [TRT] *************** Autotuning Reformat:Half(100352,784:2,28,1) -> Half(200704,784,28,1) *************** [TRT] *************** Autotuning Reformat:Float(137984,784,28,1) -> Float(200704,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.149453 [TRT] Tactic: 0 Time: 0.0358595 [TRT] Fastest Tactic: 0 Time: 0.0358595 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(137984,784,28,1) -> Float(200704,1,7168,256) *************** [TRT] --------------- Timing Runner: inception_3a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.086146 [TRT] Tactic: 0 Time: 0.115364 [TRT] Fastest Tactic: 1002 Time: 0.086146 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(137984,784,28,1) -> Half(200704,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.148933 [TRT] Tactic: 0 Time: 0.108333 [TRT] Fastest Tactic: 0 Time: 0.108333 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(137984,784,28,1) -> Half(100352,784:2,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.11823 [TRT] Tactic: 0 Time: 0.128698 [TRT] Fastest Tactic: 1002 Time: 0.11823 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(137984,1,4928,176) -> Float(200704,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.112526 [TRT] Tactic: 0 Time: 0.118177 [TRT] Fastest Tactic: 1002 Time: 0.112526 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(137984,1,4928,176) -> Float(200704,1,7168,256) *************** [TRT] --------------- Timing Runner: inception_3a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.164922 [TRT] Tactic: 0 Time: 0.115183 [TRT] Fastest Tactic: 0 Time: 0.115183 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(137984,1,4928,176) -> Half(200704,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0783595 [TRT] Tactic: 0 Time: 0.11388 [TRT] Fastest Tactic: 1002 Time: 0.0783595 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(137984,1,4928,176) -> Half(100352,784:2,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.106067 [TRT] Tactic: 0 Time: 0.13138 [TRT] Fastest Tactic: 1002 Time: 0.106067 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(137984,784,28,1) -> Float(200704,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.149453 [TRT] Tactic: 0 Time: 0.108542 [TRT] Fastest Tactic: 0 Time: 0.108542 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(137984,784,28,1) -> Float(200704,1,7168,256) *************** [TRT] --------------- Timing Runner: inception_3a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.074635 [TRT] Tactic: 0 Time: 0.116953 [TRT] Fastest Tactic: 1002 Time: 0.074635 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(137984,784,28,1) -> Half(200704,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.102396 [TRT] Tactic: 0 Time: 0.0154685 [TRT] Fastest Tactic: 0 Time: 0.0154685 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(137984,784,28,1) -> Half(100352,784:2,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0752345 [TRT] Tactic: 0 Time: 0.0647395 [TRT] Fastest Tactic: 0 Time: 0.0647395 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(68992,784:2,28,1) -> Float(200704,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.100416 [TRT] Tactic: 0 Time: 0.125782 [TRT] Fastest Tactic: 1002 Time: 0.100416 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(68992,784:2,28,1) -> Float(200704,1,7168,256) *************** [TRT] --------------- Timing Runner: inception_3a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0741145 [TRT] Tactic: 0 Time: 0.135807 [TRT] Fastest Tactic: 1002 Time: 0.0741145 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(68992,784:2,28,1) -> Half(200704,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.144427 [TRT] Tactic: 0 Time: 0.114063 [TRT] Fastest Tactic: 0 Time: 0.114063 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(68992,784:2,28,1) -> Half(100352,784:2,28,1) *************** [TRT] --------------- Timing Runner: inception_3a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.11974 [TRT] Tactic: 0 Time: 0.014297 [TRT] Fastest Tactic: 0 Time: 0.014297 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(200704,784,28,1) -> Float(200704,1,7168,256) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.306693 [TRT] Tactic: 0 Time: 0.443568 [TRT] Fastest Tactic: 1002 Time: 0.306693 [TRT] *************** Autotuning Reformat:Float(200704,784,28,1) -> Half(200704,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.482318 [TRT] Tactic: 0 Time: 0.309141 [TRT] Fastest Tactic: 0 Time: 0.309141 [TRT] *************** Autotuning Reformat:Float(200704,784,28,1) -> Half(100352,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.438698 [TRT] Tactic: 0 Time: 0.246042 [TRT] Fastest Tactic: 0 Time: 0.246042 [TRT] *************** Autotuning Reformat:Float(200704,1,7168,256) -> Float(200704,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.371615 [TRT] Tactic: 0 Time: 0.43474 [TRT] Fastest Tactic: 1002 Time: 0.371615 [TRT] *************** Autotuning Reformat:Float(200704,1,7168,256) -> Half(200704,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.290599 [TRT] Tactic: 0 Time: 0.432943 [TRT] Fastest Tactic: 1002 Time: 0.290599 [TRT] *************** Autotuning Reformat:Float(200704,1,7168,256) -> Half(100352,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.382604 [TRT] Tactic: 0 Time: 0.499818 [TRT] Fastest Tactic: 1002 Time: 0.382604 [TRT] *************** Autotuning Reformat:Half(200704,784,28,1) -> Float(200704,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.488333 [TRT] Tactic: 0 Time: 0.262917 [TRT] Fastest Tactic: 0 Time: 0.262917 [TRT] *************** Autotuning Reformat:Half(200704,784,28,1) -> Float(200704,1,7168,256) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.274115 [TRT] Tactic: 0 Time: 0.450157 [TRT] Fastest Tactic: 1002 Time: 0.274115 [TRT] *************** Autotuning Reformat:Half(200704,784,28,1) -> Half(100352,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.277579 [TRT] Tactic: 0 Time: 0.243881 [TRT] Fastest Tactic: 0 Time: 0.243881 [TRT] *************** Autotuning Reformat:Half(100352,784:2,28,1) -> Float(200704,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.36138 [TRT] Tactic: 0 Time: 0.210599 [TRT] Fastest Tactic: 0 Time: 0.210599 [TRT] *************** Autotuning Reformat:Half(100352,784:2,28,1) -> Float(200704,1,7168,256) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.276823 [TRT] Tactic: 0 Time: 0.523672 [TRT] Fastest Tactic: 1002 Time: 0.276823 [TRT] *************** Autotuning Reformat:Half(100352,784:2,28,1) -> Half(200704,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.516328 [TRT] Tactic: 0 Time: 0.207188 [TRT] Fastest Tactic: 0 Time: 0.207188 [TRT] *************** Autotuning format combination: Float(200704,784,28,1) -> Float(225792,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce (FusedConvActConvolution) [TRT] FusedConvActConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce (CudnnConvolution) [TRT] Tactic: 0 Time: 2.2756 [TRT] Tactic: 1 Time: 1.9013 [TRT] Tactic: 2 Time: 2.34549 [TRT] Tactic: 4 skipped. Scratch requested: 644251648, available: 33554432 [TRT] Tactic: 5 Time: 8.50542 [TRT] Fastest Tactic: 1 Time: 1.9013 [TRT] --------------- Timing Runner: inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce (CaskConvolution) [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.993672 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v0 Tactic: 1698681053543049347 [TRT] Tactic: 1698681053543049347 Time: 0.918515 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.99237 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.854375 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 1.08341 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v1 Tactic: 5326823351883942011 [TRT] Tactic: 5326823351883942011 Time: 1.04367 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v0 Tactic: 5500448035057547314 [TRT] Tactic: 5500448035057547314 Time: 0.965338 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.915547 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 1.10938 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 1.11305 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v1 Tactic: -6576203419454146580 [TRT] Tactic: -6576203419454146580 Time: 0.94698 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 1.14802 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.988021 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 1.0132 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.990104 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 1.06221 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v0 Tactic: -377491875521947884 [TRT] Tactic: -377491875521947884 Time: 1.07448 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v1 Tactic: -37215280111360163 [TRT] Tactic: -37215280111360163 Time: 0.873047 [TRT] Fastest Tactic: 5137655947464784826 Time: 0.854375 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 5137655947464784826 [TRT] *************** Autotuning format combination: Float(200704,1,7168,256) -> Float(225792,1,8064,288) *************** [TRT] --------------- Timing Runner: inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce (CaskConvolution) [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 3886731678879822788 [TRT] Tactic: 3886731678879822788 Time: 0.941901 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 6629944304117643200 [TRT] Tactic: 6629944304117643200 Time: 1.55016 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 1.5756 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.92724 [TRT] Fastest Tactic: -7394439838318485025 Time: 0.92724 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(200704,784,28,1) -> Half(225792,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce (CudnnConvolution) [TRT] Tactic: 0 Time: 1.71617 [TRT] Tactic: 1 Time: 1.49883 [TRT] Tactic: 2 Time: 1.70987 [TRT] Tactic: 4 skipped. Scratch requested: 644251648, available: 33554432 [TRT] Tactic: 5 Time: 6.21612 [TRT] Fastest Tactic: 1 Time: 1.49883 [TRT] --------------- Timing Runner: inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 1 [TRT] *************** Autotuning format combination: Half(100352,784:2,28,1) -> Half(225792,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] *************** Autotuning format combination: Half(100352,784:2,28,1) -> Half(112896,784:2,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce (FusedConvActConvolution) [TRT] FusedConvActConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce (CaskConvolution) [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_interior_nn_v1 Tactic: 3066127711859985668 [TRT] Tactic: 3066127711859985668 Time: 0.473203 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.52901 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.493386 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.445442 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] Tactic: 8163473458334948789 Time: 0.427239 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.508646 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.518672 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.432968 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_interior_nn_v1 Tactic: -1716393687483585322 [TRT] Tactic: -1716393687483585322 Time: 0.506562 [TRT] Fastest Tactic: 8163473458334948789 Time: 0.427239 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 8163473458334948789 [TRT] *************** Autotuning Reformat:Float(225792,784,28,1) -> Float(225792,1,8064,288) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.274792 [TRT] Tactic: 0 Time: 0.384088 [TRT] Fastest Tactic: 1002 Time: 0.274792 [TRT] *************** Autotuning Reformat:Float(225792,784,28,1) -> Half(225792,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.364713 [TRT] Tactic: 0 Time: 0.235573 [TRT] Fastest Tactic: 0 Time: 0.235573 [TRT] *************** Autotuning Reformat:Float(225792,784,28,1) -> Half(112896,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.332682 [TRT] Tactic: 0 Time: 0.1875 [TRT] Fastest Tactic: 0 Time: 0.1875 [TRT] *************** Autotuning Reformat:Float(225792,1,8064,288) -> Float(225792,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.312578 [TRT] Tactic: 0 Time: 0.364532 [TRT] Fastest Tactic: 1002 Time: 0.312578 [TRT] *************** Autotuning Reformat:Float(225792,1,8064,288) -> Half(225792,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.220338 [TRT] Tactic: 0 Time: 0.359297 [TRT] Fastest Tactic: 1002 Time: 0.220338 [TRT] *************** Autotuning Reformat:Float(225792,1,8064,288) -> Half(112896,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.290339 [TRT] Tactic: 0 Time: 0.4075 [TRT] Fastest Tactic: 1002 Time: 0.290339 [TRT] *************** Autotuning Reformat:Half(225792,784,28,1) -> Float(225792,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.369166 [TRT] Tactic: 0 Time: 0.19987 [TRT] Fastest Tactic: 0 Time: 0.19987 [TRT] *************** Autotuning Reformat:Half(225792,784,28,1) -> Float(225792,1,8064,288) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.208072 [TRT] Tactic: 0 Time: 0.342891 [TRT] Fastest Tactic: 1002 Time: 0.208072 [TRT] *************** Autotuning Reformat:Half(225792,784,28,1) -> Half(112896,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.218932 [TRT] Tactic: 0 Time: 0.186302 [TRT] Fastest Tactic: 0 Time: 0.186302 [TRT] *************** Autotuning Reformat:Half(112896,784:2,28,1) -> Float(225792,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.273568 [TRT] Tactic: 0 Time: 0.163125 [TRT] Fastest Tactic: 0 Time: 0.163125 [TRT] *************** Autotuning Reformat:Half(112896,784:2,28,1) -> Float(225792,1,8064,288) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.209948 [TRT] Tactic: 0 Time: 0.396484 [TRT] Fastest Tactic: 1002 Time: 0.209948 [TRT] *************** Autotuning Reformat:Half(112896,784:2,28,1) -> Half(225792,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.421745 [TRT] Tactic: 0 Time: 0.158698 [TRT] Fastest Tactic: 0 Time: 0.158698 [TRT] *************** Autotuning Reformat:Float(225792,784,28,1) -> Float(225792,1,8064,288) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.113307 [TRT] Tactic: 0 Time: 0.170365 [TRT] Fastest Tactic: 1002 Time: 0.113307 [TRT] *************** Autotuning Reformat:Float(225792,784,28,1) -> Half(225792,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.196094 [TRT] Tactic: 0 Time: 0.143412 [TRT] Fastest Tactic: 0 Time: 0.143412 [TRT] *************** Autotuning Reformat:Float(225792,784,28,1) -> Half(112896,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.154583 [TRT] Tactic: 0 Time: 0.170026 [TRT] Fastest Tactic: 1002 Time: 0.154583 [TRT] *************** Autotuning Reformat:Float(225792,1,8064,288) -> Float(225792,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.140417 [TRT] Tactic: 0 Time: 0.16651 [TRT] Fastest Tactic: 1002 Time: 0.140417 [TRT] *************** Autotuning Reformat:Float(225792,1,8064,288) -> Half(225792,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.102031 [TRT] Tactic: 0 Time: 0.164636 [TRT] Fastest Tactic: 1002 Time: 0.102031 [TRT] *************** Autotuning Reformat:Float(225792,1,8064,288) -> Half(112896,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.135443 [TRT] Tactic: 0 Time: 0.186458 [TRT] Fastest Tactic: 1002 Time: 0.135443 [TRT] *************** Autotuning Reformat:Half(225792,784,28,1) -> Float(225792,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.198438 [TRT] Tactic: 0 Time: 0.092396 [TRT] Fastest Tactic: 0 Time: 0.092396 [TRT] *************** Autotuning Reformat:Half(225792,784,28,1) -> Float(225792,1,8064,288) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.096172 [TRT] Tactic: 0 Time: 0.155782 [TRT] Fastest Tactic: 1002 Time: 0.096172 [TRT] *************** Autotuning Reformat:Half(225792,784,28,1) -> Half(112896,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.103984 [TRT] Tactic: 0 Time: 0.0858075 [TRT] Fastest Tactic: 0 Time: 0.0858075 [TRT] *************** Autotuning Reformat:Half(112896,784:2,28,1) -> Float(225792,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.128385 [TRT] Tactic: 0 Time: 0.075234 [TRT] Fastest Tactic: 0 Time: 0.075234 [TRT] *************** Autotuning Reformat:Half(112896,784:2,28,1) -> Float(225792,1,8064,288) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0961725 [TRT] Tactic: 0 Time: 0.179271 [TRT] Fastest Tactic: 1002 Time: 0.0961725 [TRT] *************** Autotuning Reformat:Half(112896,784:2,28,1) -> Half(225792,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.182786 [TRT] Tactic: 0 Time: 0.151276 [TRT] Fastest Tactic: 0 Time: 0.151276 [TRT] *************** Autotuning format combination: Float(225792,784,28,1) -> Float(376320,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/3x3 + inception_3b/relu_3x3 (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/3x3 + inception_3b/relu_3x3 (FusedConvActConvolution) [TRT] FusedConvActConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/3x3 + inception_3b/relu_3x3 (CudnnConvolution) [TRT] Tactic: 0 Time: 4.49951 [TRT] Tactic: 1 Time: 2.30857 [TRT] Tactic: 2 Time: 4.02456 [TRT] Tactic: 4 skipped. Scratch requested: 215908352, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 108347392, available: 33554432 [TRT] Tactic: 6 Time: 1.68326 [TRT] Fastest Tactic: 6 Time: 1.68326 [TRT] --------------- Timing Runner: inception_3b/3x3 + inception_3b/relu_3x3 (CaskConvolution) [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 2.55031 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v0 Tactic: 1754984623894446479 [TRT] Tactic: 1754984623894446479 Time: 3.00221 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v0 Tactic: 3611739942397549984 [TRT] Tactic: 3611739942397549984 Time: 2.76409 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148n_nt_v1 Tactic: 3827454225649558724 [TRT] Tactic: 3827454225649558724 Time: 2.01174 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v1 Tactic: 4337000649858996379 [TRT] Tactic: 4337000649858996379 Time: 2.13651 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 2.73398 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 2.01101 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 2.68091 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 5921334924264294896 [TRT] Tactic: 5921334924264294896 Time: 1.49982 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 2.11253 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 2.65914 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148t_nt_v1 Tactic: 7852627285308570038 [TRT] Tactic: 7852627285308570038 Time: 2.01716 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v1 Tactic: -9137461792520977713 [TRT] Tactic: -9137461792520977713 Time: 2.74776 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148n_nt_v0 Tactic: -8776506421218919509 [TRT] Tactic: -8776506421218919509 Time: 1.95914 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 2.74021 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v0 Tactic: -8133971918129952780 [TRT] Tactic: -8133971918129952780 Time: 2.35161 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v1 Tactic: -6092040395344634144 [TRT] Tactic: -6092040395344634144 Time: 2.63742 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 3.01591 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 2.34719 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148t_nt_v0 Tactic: -2318106587342035239 [TRT] Tactic: -2318106587342035239 Time: 2.02255 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_mobile_relu_tile148t_nt_v0 Tactic: -1343271414618805657 [TRT] Tactic: -1343271414618805657 Time: 1.36951 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 2.34826 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 2.28182 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 2.63109 [TRT] Fastest Tactic: -1343271414618805657 Time: 1.36951 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -1343271414618805657 [TRT] *************** Autotuning format combination: Float(225792,1,8064,288) -> Float(376320,1,13440,480) *************** [TRT] --------------- Timing Runner: inception_3b/3x3 + inception_3b/relu_3x3 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/3x3 + inception_3b/relu_3x3 (CaskConvolution) [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 2.56883 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 2.00732 [TRT] Fastest Tactic: -7394439838318485025 Time: 2.00732 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(225792,784,28,1) -> Half(376320,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/3x3 + inception_3b/relu_3x3 (CudnnConvolution) [TRT] Tactic: 0 Time: 4.36227 [TRT] Tactic: 1 Time: 4.43341 [TRT] Tactic: 2 Time: 3.87047 [TRT] Tactic: 4 skipped. Scratch requested: 215908352, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 108347392, available: 33554432 [TRT] Tactic: 6 Time: 5.49971 [TRT] Fastest Tactic: 2 Time: 3.87047 [TRT] --------------- Timing Runner: inception_3b/3x3 + inception_3b/relu_3x3 (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 2 [TRT] *************** Autotuning format combination: Half(112896,784:2,28,1) -> Half(188160,784:2,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/3x3 + inception_3b/relu_3x3 (FusedConvActConvolution) [TRT] FusedConvActConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/3x3 + inception_3b/relu_3x3 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/3x3 + inception_3b/relu_3x3 (CaskConvolution) [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 1.34185 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_large_nn_v1 Tactic: 3650389455493082349 [TRT] Tactic: 3650389455493082349 Time: 1.35073 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_winograd_fp16x2_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 4772821744921268633 [TRT] Tactic: 4772821744921268633 Time: 0.79724 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 1.16589 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 1.0506 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_large_nn_v1 Tactic: -6490690591794140522 [TRT] Tactic: -6490690591794140522 Time: 1.06258 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_large_nn_v1 Tactic: -4686027666808657977 [TRT] Tactic: -4686027666808657977 Time: 1.38865 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 1.33021 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 1.38427 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 1.02612 [TRT] Fastest Tactic: 4772821744921268633 Time: 0.79724 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 4772821744921268633 [TRT] *************** Autotuning Reformat:Float(376320,784,28,1) -> Float(376320,1,13440,480) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.166068 [TRT] Tactic: 0 Time: 0.245651 [TRT] Fastest Tactic: 1002 Time: 0.166068 [TRT] *************** Autotuning Reformat:Float(376320,784,28,1) -> Half(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.291354 [TRT] Tactic: 0 Time: 0.210833 [TRT] Fastest Tactic: 0 Time: 0.210833 [TRT] *************** Autotuning Reformat:Float(376320,784,28,1) -> Half(188160,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.224791 [TRT] Tactic: 0 Time: 0.12724 [TRT] Fastest Tactic: 0 Time: 0.12724 [TRT] *************** Autotuning Reformat:Float(376320,1,13440,480) -> Float(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.200678 [TRT] Tactic: 0 Time: 0.242786 [TRT] Fastest Tactic: 1002 Time: 0.200678 [TRT] *************** Autotuning Reformat:Float(376320,1,13440,480) -> Half(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.150053 [TRT] Tactic: 0 Time: 0.239219 [TRT] Fastest Tactic: 1002 Time: 0.150053 [TRT] *************** Autotuning Reformat:Float(376320,1,13440,480) -> Half(188160,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.197266 [TRT] Tactic: 0 Time: 0.270599 [TRT] Fastest Tactic: 1002 Time: 0.197266 [TRT] *************** Autotuning Reformat:Half(376320,784,28,1) -> Float(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.294062 [TRT] Tactic: 0 Time: 0.211979 [TRT] Fastest Tactic: 0 Time: 0.211979 [TRT] *************** Autotuning Reformat:Half(376320,784,28,1) -> Float(376320,1,13440,480) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.140651 [TRT] Tactic: 0 Time: 0.230469 [TRT] Fastest Tactic: 1002 Time: 0.140651 [TRT] *************** Autotuning Reformat:Half(376320,784,28,1) -> Half(188160,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.14948 [TRT] Tactic: 0 Time: 0.126614 [TRT] Fastest Tactic: 0 Time: 0.126614 [TRT] *************** Autotuning Reformat:Half(188160,784:2,28,1) -> Float(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.186563 [TRT] Tactic: 0 Time: 0.248386 [TRT] Fastest Tactic: 1002 Time: 0.186563 [TRT] *************** Autotuning Reformat:Half(188160,784:2,28,1) -> Float(376320,1,13440,480) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.142343 [TRT] Tactic: 0 Time: 0.266693 [TRT] Fastest Tactic: 1002 Time: 0.142343 [TRT] *************** Autotuning Reformat:Half(188160,784:2,28,1) -> Half(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.277266 [TRT] Tactic: 0 Time: 0.223958 [TRT] Fastest Tactic: 0 Time: 0.223958 [TRT] *************** Autotuning Reformat:Float(225792,784,28,1) -> Float(225792,1,8064,288) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0602085 [TRT] Tactic: 0 Time: 0.0412505 [TRT] Fastest Tactic: 0 Time: 0.0412505 [TRT] *************** Autotuning Reformat:Float(225792,784,28,1) -> Half(225792,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0529685 [TRT] Tactic: 0 Time: 0.038203 [TRT] Fastest Tactic: 0 Time: 0.038203 [TRT] *************** Autotuning Reformat:Float(225792,784,28,1) -> Half(112896,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.044922 [TRT] Tactic: 0 Time: 0.044896 [TRT] Fastest Tactic: 0 Time: 0.044896 [TRT] *************** Autotuning Reformat:Float(225792,1,8064,288) -> Float(225792,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.063542 [TRT] Tactic: 0 Time: 0.038386 [TRT] Fastest Tactic: 0 Time: 0.038386 [TRT] *************** Autotuning Reformat:Float(225792,1,8064,288) -> Half(225792,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.028854 [TRT] Tactic: 0 Time: 0.0382815 [TRT] Fastest Tactic: 1002 Time: 0.028854 [TRT] *************** Autotuning Reformat:Float(225792,1,8064,288) -> Half(112896,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0403385 [TRT] Tactic: 0 Time: 0.0456505 [TRT] Fastest Tactic: 1002 Time: 0.0403385 [TRT] *************** Autotuning Reformat:Half(225792,784,28,1) -> Float(225792,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0540105 [TRT] Tactic: 0 Time: 0.0264325 [TRT] Fastest Tactic: 0 Time: 0.0264325 [TRT] *************** Autotuning Reformat:Half(225792,784,28,1) -> Float(225792,1,8064,288) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0278385 [TRT] Tactic: 0 Time: 0.042813 [TRT] Fastest Tactic: 1002 Time: 0.0278385 [TRT] *************** Autotuning Reformat:Half(225792,784,28,1) -> Half(112896,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.028776 [TRT] Tactic: 0 Time: 0.024062 [TRT] Fastest Tactic: 0 Time: 0.024062 [TRT] *************** Autotuning Reformat:Half(112896,784:2,28,1) -> Float(225792,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.038099 [TRT] Tactic: 0 Time: 0.0217965 [TRT] Fastest Tactic: 0 Time: 0.0217965 [TRT] *************** Autotuning Reformat:Half(112896,784:2,28,1) -> Float(225792,1,8064,288) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.028125 [TRT] Tactic: 0 Time: 0.047526 [TRT] Fastest Tactic: 1002 Time: 0.028125 [TRT] *************** Autotuning Reformat:Half(112896,784:2,28,1) -> Half(225792,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.075182 [TRT] Tactic: 0 Time: 0.040339 [TRT] Fastest Tactic: 0 Time: 0.040339 [TRT] *************** Autotuning format combination: Float(225792,784,28,1) -> Float(376320,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/5x5 + inception_3b/relu_5x5 (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/5x5 + inception_3b/relu_5x5 (FusedConvActConvolution) [TRT] Tactic: 393215 Time: 0.801432 [TRT] Tactic: 917503 Time: 0.836614 [TRT] Tactic: 1114111 Time: 1.03396 [TRT] Tactic: 1245183 Time: 1.01971 [TRT] Tactic: 1572863 Time: 0.809349 [TRT] Tactic: 2490367 Time: 0.887031 [TRT] Tactic: 2555903 Time: 0.902812 [TRT] Tactic: 2949119 Time: 1.25966 [TRT] Tactic: 3211263 Time: 1.14174 [TRT] Tactic: 3801087 Time: 1.05021 [TRT] Tactic: 3866623 Time: 0.870157 [TRT] Tactic: 4128767 Time: 1.10659 [TRT] Tactic: 4456447 Time: 0.850834 [TRT] Tactic: 4718591 Time: 1.13424 [TRT] Tactic: 4784127 Time: 1.11568 [TRT] Tactic: 4849663 Time: 1.00641 [TRT] Tactic: 5111807 Time: 1.0594 [TRT] Tactic: 5308415 Time: 1.12096 [TRT] Tactic: 5505023 Time: 1.24391 [TRT] Tactic: 6094847 Time: 0.991719 [TRT] Tactic: 6356991 Time: 0.912396 [TRT] Tactic: 6553599 Time: 0.802136 [TRT] Tactic: 6619135 Time: 0.984506 [TRT] Tactic: 6684671 Time: 1.2407 [TRT] Tactic: 7471103 Time: 0.838906 [TRT] Tactic: 7667711 Time: 1.11352 [TRT] Tactic: 7929855 Time: 1.47107 [TRT] Tactic: 8060927 Time: 1.00414 [TRT] Tactic: 8126463 Time: 1.36758 [TRT] Tactic: 8388607 Time: 1.17174 [TRT] Tactic: 8519679 Time: 1.33862 [TRT] Tactic: 8781823 Time: 1.35385 [TRT] Tactic: 8912895 Time: 1.38521 [TRT] Tactic: 9240575 Time: 1.37164 [TRT] Tactic: 9306111 Time: 1.31505 [TRT] Tactic: 9371647 Time: 1.06115 [TRT] Tactic: 9437183 Time: 1.25774 [TRT] Tactic: 9633791 Time: 1.00589 [TRT] Tactic: 9699327 Time: 0.925573 [TRT] Tactic: 9764863 Time: 0.803099 [TRT] Tactic: 10158079 Time: 0.858828 [TRT] Tactic: 10420223 Time: 1.32018 [TRT] Tactic: 10616831 Time: 1.06643 [TRT] Tactic: 10878975 Time: 0.814089 [TRT] Fastest Tactic: 393215 Time: 0.801432 [TRT] --------------- Timing Runner: inception_3b/5x5 + inception_3b/relu_5x5 (CudnnConvolution) [TRT] Tactic: 0 Time: 1.65648 [TRT] Tactic: 1 Time: 0.938255 [TRT] Tactic: 2 Time: 1.56776 [TRT] Tactic: 4 Time: 4.44888 [TRT] Tactic: 5 Time: 4.59055 [TRT] Fastest Tactic: 1 Time: 0.938255 [TRT] --------------- Timing Runner: inception_3b/5x5 + inception_3b/relu_5x5 (CaskConvolution) [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 1.00904 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v0 Tactic: 1754984623894446479 [TRT] Tactic: 1754984623894446479 Time: 1.19781 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v0 Tactic: 3611739942397549984 [TRT] Tactic: 3611739942397549984 Time: 1.02065 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v1 Tactic: 4337000649858996379 [TRT] Tactic: 4337000649858996379 Time: 1.05393 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 1.01099 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.936875 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.965756 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 1.03825 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.929584 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v1 Tactic: -9137461792520977713 [TRT] Tactic: -9137461792520977713 Time: 1.02112 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 1.02667 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v0 Tactic: -8133971918129952780 [TRT] Tactic: -8133971918129952780 Time: 1.10029 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v1 Tactic: -6092040395344634144 [TRT] Tactic: -6092040395344634144 Time: 1.05995 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 1.18737 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.849272 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 1.09292 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 1.02562 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.957735 [TRT] Fastest Tactic: -3456450830548107839 Time: 0.849272 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 393215 [TRT] *************** Autotuning format combination: Float(225792,1,8064,288) -> Float(376320,1,13440,480) *************** [TRT] --------------- Timing Runner: inception_3b/5x5 + inception_3b/relu_5x5 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/5x5 + inception_3b/relu_5x5 (CaskConvolution) [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.921302 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.945104 [TRT] Fastest Tactic: -9153228964338181824 Time: 0.921302 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -9153228964338181824 [TRT] *************** Autotuning format combination: Half(225792,784,28,1) -> Half(376320,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/5x5 + inception_3b/relu_5x5 (CudnnConvolution) [TRT] Tactic: 0 Time: 1.59109 [TRT] Tactic: 1 Time: 1.59435 [TRT] Tactic: 2 Time: 1.7137 [TRT] Tactic: 4 Time: 4.49266 [TRT] Tactic: 5 Time: 4.53133 [TRT] Fastest Tactic: 0 Time: 1.59109 [TRT] --------------- Timing Runner: inception_3b/5x5 + inception_3b/relu_5x5 (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 0 [TRT] *************** Autotuning format combination: Half(112896,784:2,28,1) -> Half(188160,784:2,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/5x5 + inception_3b/relu_5x5 (FusedConvActConvolution) [TRT] Tactic: 393215 Time: 0.40237 [TRT] Tactic: 917503 Time: 0.46599 [TRT] Tactic: 1114111 Time: 0.518281 [TRT] Tactic: 1245183 Time: 0.556615 [TRT] Tactic: 1572863 Time: 0.455547 [TRT] Tactic: 2490367 Time: 0.522395 [TRT] Tactic: 2555903 Time: 0.571537 [TRT] Tactic: 2949119 Time: 0.62362 [TRT] Tactic: 3211263 Time: 0.611562 [TRT] Tactic: 3801087 Time: 0.496615 [TRT] Tactic: 3866623 Time: 0.443516 [TRT] Tactic: 4128767 Time: 0.591223 [TRT] Tactic: 4456447 Time: 0.472448 [TRT] Tactic: 4718591 Time: 0.522891 [TRT] Tactic: 4784127 Time: 0.614714 [TRT] Tactic: 4849663 Time: 0.507474 [TRT] Tactic: 5111807 Time: 0.51901 [TRT] Tactic: 5308415 Time: 0.571536 [TRT] Tactic: 5505023 Time: 0.595651 [TRT] Tactic: 6094847 Time: 0.494166 [TRT] Tactic: 6356991 Time: 0.557813 [TRT] Tactic: 6553599 Time: 0.402448 [TRT] Tactic: 6619135 Time: 0.47388 [TRT] Tactic: 6684671 Time: 0.560313 [TRT] Tactic: 7471103 Time: 0.444427 [TRT] Tactic: 7667711 Time: 0.52211 [TRT] Tactic: 7929855 Time: 0.527682 [TRT] Tactic: 8060927 Time: 0.465756 [TRT] Tactic: 8126463 Time: 0.579427 [TRT] Tactic: 8388607 Time: 0.603724 [TRT] Tactic: 8519679 Time: 0.520105 [TRT] Tactic: 8781823 Time: 0.676485 [TRT] Tactic: 8912895 Time: 0.665756 [TRT] Tactic: 9240575 Time: 0.533854 [TRT] Tactic: 9306111 Time: 0.637943 [TRT] Tactic: 9371647 Time: 0.513516 [TRT] Tactic: 9437183 Time: 0.622057 [TRT] Tactic: 9633791 Time: 0.506745 [TRT] Tactic: 9699327 Time: 0.424688 [TRT] Tactic: 9764863 Time: 0.432943 [TRT] Tactic: 10158079 Time: 0.426562 [TRT] Tactic: 10420223 Time: 0.630937 [TRT] Tactic: 10616831 Time: 0.496198 [TRT] Tactic: 10878975 Time: 0.442838 [TRT] Fastest Tactic: 393215 Time: 0.40237 [TRT] --------------- Timing Runner: inception_3b/5x5 + inception_3b/relu_5x5 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/5x5 + inception_3b/relu_5x5 (CaskConvolution) [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.525391 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_large_nn_v1 Tactic: 3650389455493082349 [TRT] Tactic: 3650389455493082349 Time: 0.547943 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.440702 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.527005 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_large_nn_v1 Tactic: -6490690591794140522 [TRT] Tactic: -6490690591794140522 Time: 0.530469 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_large_nn_v1 Tactic: -4686027666808657977 [TRT] Tactic: -4686027666808657977 Time: 0.516354 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.474739 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.512422 [TRT] inception_3b/5x5 + inception_3b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.484375 [TRT] Fastest Tactic: 5319956359050645452 Time: 0.440702 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 393215 [TRT] *************** Autotuning Reformat:Float(376320,784,28,1) -> Float(376320,1,13440,480) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.115729 [TRT] Tactic: 0 Time: 0.128516 [TRT] Fastest Tactic: 1002 Time: 0.115729 [TRT] *************** Autotuning Reformat:Float(376320,784,28,1) -> Half(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.148933 [TRT] Tactic: 0 Time: 0.108047 [TRT] Fastest Tactic: 0 Time: 0.108047 [TRT] *************** Autotuning Reformat:Float(376320,784,28,1) -> Half(188160,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.117084 [TRT] Tactic: 0 Time: 0.065625 [TRT] Fastest Tactic: 0 Time: 0.065625 [TRT] *************** Autotuning Reformat:Float(376320,1,13440,480) -> Float(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.131719 [TRT] Tactic: 0 Time: 0.123437 [TRT] Fastest Tactic: 0 Time: 0.123437 [TRT] *************** Autotuning Reformat:Float(376320,1,13440,480) -> Half(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.077839 [TRT] Tactic: 0 Time: 0.122266 [TRT] Fastest Tactic: 1002 Time: 0.077839 [TRT] *************** Autotuning Reformat:Float(376320,1,13440,480) -> Half(188160,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.10362 [TRT] Tactic: 0 Time: 0.138125 [TRT] Fastest Tactic: 1002 Time: 0.10362 [TRT] *************** Autotuning Reformat:Half(376320,784,28,1) -> Float(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.151432 [TRT] Tactic: 0 Time: 0.109583 [TRT] Fastest Tactic: 0 Time: 0.109583 [TRT] *************** Autotuning Reformat:Half(376320,784,28,1) -> Float(376320,1,13440,480) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.074375 [TRT] Tactic: 0 Time: 0.118437 [TRT] Fastest Tactic: 1002 Time: 0.074375 [TRT] *************** Autotuning Reformat:Half(376320,784,28,1) -> Half(188160,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.080234 [TRT] Tactic: 0 Time: 0.064922 [TRT] Fastest Tactic: 0 Time: 0.064922 [TRT] *************** Autotuning Reformat:Half(188160,784:2,28,1) -> Float(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0989325 [TRT] Tactic: 0 Time: 0.127265 [TRT] Fastest Tactic: 1002 Time: 0.0989325 [TRT] *************** Autotuning Reformat:Half(188160,784:2,28,1) -> Float(376320,1,13440,480) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.074245 [TRT] Tactic: 0 Time: 0.136771 [TRT] Fastest Tactic: 1002 Time: 0.074245 [TRT] *************** Autotuning Reformat:Half(188160,784:2,28,1) -> Half(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.143802 [TRT] Tactic: 0 Time: 0.114662 [TRT] Fastest Tactic: 0 Time: 0.114662 [TRT] *************** Autotuning Reformat:Float(200704,784,28,1) -> Half(200704,784,28,1) *************** [TRT] *************** Autotuning Reformat:Float(200704,784,28,1) -> Half(100352,784:2,28,1) *************** [TRT] *************** Autotuning Reformat:Float(200704,1,7168,256) -> Float(200704,784,28,1) *************** [TRT] *************** Autotuning Reformat:Float(200704,1,7168,256) -> Half(200704,784,28,1) *************** [TRT] *************** Autotuning Reformat:Float(200704,1,7168,256) -> Half(100352,784:2,28,1) *************** [TRT] *************** Autotuning Reformat:Half(200704,784,28,1) -> Float(200704,784,28,1) *************** [TRT] *************** Autotuning Reformat:Half(200704,784,28,1) -> Half(100352,784:2,28,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,784:2,28,1) -> Float(200704,784,28,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,784:2,28,1) -> Half(200704,784,28,1) *************** [TRT] *************** Autotuning format combination: Float(200704,784,28,1) -> Float(200704,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/pool (TiledPooling) [TRT] Tactic: 2752769 Time: 0.432709 [TRT] Tactic: 2818305 Time: 0.411303 [TRT] Tactic: 2883841 Time: 0.307891 [TRT] Tactic: 2949377 Time: 0.845156 [TRT] Tactic: 3014913 Time: 0.676614 [TRT] Tactic: 3080449 Time: 0.363932 [TRT] Tactic: 3145985 Time: 0.336666 [TRT] Tactic: 3211521 Time: 0.332943 [TRT] Tactic: 3277057 Time: 0.31875 [TRT] Tactic: 3342593 Time: 0.215287 [TRT] Tactic: 3408129 Time: 0.505339 [TRT] Tactic: 3473665 Time: 0.410755 [TRT] Tactic: 3539201 Time: 0.229792 [TRT] Tactic: 3604737 Time: 0.218516 [TRT] Tactic: 3670273 Time: 0.331562 [TRT] Tactic: 3735809 Time: 0.319271 [TRT] Tactic: 3801345 Time: 0.197161 [TRT] Tactic: 3866881 Time: 0.406797 [TRT] Tactic: 3932417 Time: 0.328021 [TRT] Tactic: 3997953 Time: 0.194818 [TRT] Tactic: 4063489 Time: 0.185391 [TRT] Tactic: 4129025 Time: 0.3325 [TRT] Tactic: 4194561 Time: 0.319922 [TRT] Tactic: 4260097 Time: 0.196432 [TRT] Tactic: 4325633 Time: 0.35164 [TRT] Tactic: 4391169 Time: 0.282188 [TRT] Tactic: 4456705 Time: 0.168672 [TRT] Tactic: 4522241 Time: 0.170574 [TRT] Tactic: 4587777 Time: 0.33487 [TRT] Tactic: 4653313 Time: 0.320521 [TRT] Tactic: 4718849 Time: 0.198438 [TRT] Tactic: 4784385 Time: 0.329922 [TRT] Tactic: 4849921 Time: 0.269895 [TRT] Tactic: 4915457 Time: 0.15914 [TRT] Tactic: 4980993 Time: 0.159818 [TRT] Tactic: 5046529 Time: 0.336693 [TRT] Tactic: 5112065 Time: 0.321198 [TRT] Tactic: 5177601 Time: 0.196875 [TRT] Tactic: 5243137 Time: 0.3075 [TRT] Tactic: 5308673 Time: 0.25086 [TRT] Tactic: 5374209 Time: 0.152031 [TRT] Tactic: 5439745 Time: 0.156614 [TRT] Tactic: 6553857 Time: 0.145938 [TRT] Tactic: 6750465 Time: 0.216172 [TRT] Fastest Tactic: 6553857 Time: 0.145938 [TRT] --------------- Timing Runner: inception_3b/pool (CudnnPooling) [TRT] Tactic: -1 Time: 0.603256 [TRT] Fastest Tactic: -1 Time: 0.603256 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: TiledPooling Tactic: 6553857 [TRT] *************** Autotuning format combination: Half(200704,784,28,1) -> Half(200704,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/pool (TiledPooling) [TRT] TiledPooling has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/pool (CudnnPooling) [TRT] Tactic: -1 Time: 0.632448 [TRT] Fastest Tactic: -1 Time: 0.632448 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnPooling Tactic: -1 [TRT] *************** Autotuning format combination: Half(100352,784:2,28,1) -> Half(100352,784:2,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/pool (TiledPooling) [TRT] Tactic: 2752769 Time: 0.238621 [TRT] Tactic: 2818305 Time: 0.234062 [TRT] Tactic: 2883841 Time: 0.179765 [TRT] Tactic: 2949377 Time: 0.462266 [TRT] Tactic: 3014913 Time: 0.382553 [TRT] Tactic: 3080449 Time: 0.210443 [TRT] Tactic: 3145985 Time: 0.189844 [TRT] Tactic: 3211521 Time: 0.170911 [TRT] Tactic: 3277057 Time: 0.16526 [TRT] Tactic: 3342593 Time: 0.133802 [TRT] Tactic: 3408129 Time: 0.313359 [TRT] Tactic: 3473665 Time: 0.255729 [TRT] Tactic: 3539201 Time: 0.142969 [TRT] Tactic: 3604737 Time: 0.136198 [TRT] Tactic: 3670273 Time: 0.157943 [TRT] Tactic: 3735809 Time: 0.154323 [TRT] Tactic: 3801345 Time: 0.113984 [TRT] Tactic: 3866881 Time: 0.272864 [TRT] Tactic: 3932417 Time: 0.219713 [TRT] Tactic: 3997953 Time: 0.125443 [TRT] Tactic: 4063489 Time: 0.120546 [TRT] Tactic: 4129025 Time: 0.157266 [TRT] Tactic: 4194561 Time: 0.152162 [TRT] Tactic: 4260097 Time: 0.106927 [TRT] Tactic: 4325633 Time: 0.240391 [TRT] Tactic: 4391169 Time: 0.201016 [TRT] Tactic: 4456705 Time: 0.114948 [TRT] Tactic: 4522241 Time: 0.108958 [TRT] Tactic: 4587777 Time: 0.159791 [TRT] Tactic: 4653313 Time: 0.150703 [TRT] Tactic: 4718849 Time: 0.103567 [TRT] Tactic: 4784385 Time: 0.233177 [TRT] Tactic: 4849921 Time: 0.193307 [TRT] Tactic: 4915457 Time: 0.111224 [TRT] Tactic: 4980993 Time: 0.107682 [TRT] Tactic: 5046529 Time: 0.159088 [TRT] Tactic: 5112065 Time: 0.152917 [TRT] Tactic: 5177601 Time: 0.104531 [TRT] Tactic: 5243137 Time: 0.228203 [TRT] Tactic: 5308673 Time: 0.187292 [TRT] Tactic: 5374209 Time: 0.108516 [TRT] Tactic: 5439745 Time: 0.105599 [TRT] Tactic: 6553857 Time: 0.092969 [TRT] Tactic: 6750465 Time: 0.144922 [TRT] Fastest Tactic: 6553857 Time: 0.092969 [TRT] --------------- Timing Runner: inception_3b/pool (CudaPooling) [TRT] Tactic: -3 Time: 0.362994 [TRT] Fastest Tactic: -3 Time: 0.362994 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: TiledPooling Tactic: 6553857 [TRT] *************** Autotuning Reformat:Float(200704,784,28,1) -> Float(200704,1,7168,256) *************** [TRT] *************** Autotuning Reformat:Float(200704,784,28,1) -> Half(200704,784,28,1) *************** [TRT] *************** Autotuning Reformat:Float(200704,784,28,1) -> Half(100352,784:2,28,1) *************** [TRT] *************** Autotuning Reformat:Float(200704,1,7168,256) -> Float(200704,784,28,1) *************** [TRT] *************** Autotuning Reformat:Float(200704,1,7168,256) -> Half(200704,784,28,1) *************** [TRT] *************** Autotuning Reformat:Float(200704,1,7168,256) -> Half(100352,784:2,28,1) *************** [TRT] *************** Autotuning Reformat:Half(200704,784,28,1) -> Float(200704,784,28,1) *************** [TRT] *************** Autotuning Reformat:Half(200704,784,28,1) -> Float(200704,1,7168,256) *************** [TRT] *************** Autotuning Reformat:Half(200704,784,28,1) -> Half(100352,784:2,28,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,784:2,28,1) -> Float(200704,784,28,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,784:2,28,1) -> Float(200704,1,7168,256) *************** [TRT] *************** Autotuning Reformat:Half(100352,784:2,28,1) -> Half(200704,784,28,1) *************** [TRT] *************** Autotuning format combination: Float(200704,784,28,1) -> Float(376320,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/pool_proj + inception_3b/relu_pool_proj (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/pool_proj + inception_3b/relu_pool_proj (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.376952 [TRT] Tactic: 655359 Time: 0.466015 [TRT] Tactic: 786431 Time: 0.470807 [TRT] Tactic: 851967 Time: 0.673307 [TRT] Tactic: 1179647 Time: 0.405104 [TRT] Tactic: 1310719 Time: 0.79112 [TRT] Tactic: 1376255 Time: 0.335963 [TRT] Tactic: 1441791 Time: 0.437787 [TRT] Tactic: 1507327 Time: 0.628672 [TRT] Tactic: 1638399 Time: 0.477709 [TRT] Tactic: 1835007 Time: 0.431849 [TRT] Tactic: 1900543 Time: 0.640469 [TRT] Tactic: 2097151 Time: 0.47349 [TRT] Tactic: 2162687 Time: 0.348958 [TRT] Tactic: 2293759 Time: 0.352005 [TRT] Tactic: 2359295 Time: 0.351745 [TRT] Tactic: 2686975 Time: 0.320938 [TRT] Tactic: 3080191 Time: 0.531615 [TRT] Tactic: 3342335 Time: 0.658671 [TRT] Tactic: 3407871 Time: 0.353985 [TRT] Tactic: 3538943 Time: 0.351562 [TRT] Tactic: 3670015 Time: 0.430286 [TRT] Tactic: 3932159 Time: 0.594895 [TRT] Tactic: 3997695 Time: 0.476849 [TRT] Tactic: 4063231 Time: 0.631328 [TRT] Tactic: 4194303 Time: 0.399922 [TRT] Tactic: 4259839 Time: 0.48388 [TRT] Tactic: 4325375 Time: 0.434192 [TRT] Tactic: 4521983 Time: 0.464193 [TRT] Tactic: 4587519 Time: 0.498671 [TRT] Tactic: 4653055 Time: 0.460156 [TRT] Tactic: 4915199 Time: 0.409088 [TRT] Tactic: 4980735 Time: 0.440755 [TRT] Tactic: 5177343 Time: 0.45888 [TRT] Tactic: 5242879 Time: 0.334219 [TRT] Tactic: 5373951 Time: 0.581692 [TRT] Tactic: 5439487 Time: 0.60151 [TRT] Tactic: 5570559 Time: 0.437448 [TRT] Tactic: 5636095 Time: 0.628905 [TRT] Tactic: 5701631 Time: 0.387865 [TRT] Tactic: 5767167 Time: 1.1499 [TRT] Tactic: 5832703 Time: 0.353047 [TRT] Tactic: 5898239 Time: 0.319192 [TRT] Tactic: 6029311 Time: 0.332839 [TRT] Tactic: 6225919 Time: 0.32888 [TRT] Tactic: 6291455 Time: 0.404401 [TRT] Tactic: 6422527 Time: 0.535417 [TRT] Tactic: 6750207 Time: 0.430573 [TRT] Tactic: 6815743 Time: 0.409427 [TRT] Tactic: 6946815 Time: 0.645157 [TRT] Tactic: 7012351 Time: 0.474375 [TRT] Tactic: 7077887 Time: 0.349349 [TRT] Tactic: 7143423 Time: 0.664089 [TRT] Tactic: 7208959 Time: 0.348281 [TRT] Tactic: 7340031 Time: 0.35513 [TRT] Tactic: 7405567 Time: 0.382839 [TRT] Tactic: 7536639 Time: 0.397709 [TRT] Tactic: 7602175 Time: 0.519662 [TRT] Tactic: 7733247 Time: 0.361328 [TRT] Tactic: 7798783 Time: 0.467135 [TRT] Tactic: 8191999 Time: 0.648593 [TRT] Tactic: 8257535 Time: 0.422526 [TRT] Tactic: 8323071 Time: 0.432136 [TRT] Tactic: 8650751 Time: 0.523177 [TRT] Tactic: 8716287 Time: 0.444896 [TRT] Tactic: 9109503 Time: 0.539558 [TRT] Tactic: 9568255 Time: 0.415 [TRT] Tactic: 9895935 Time: 0.399219 [TRT] Tactic: 10223615 Time: 0.321302 [TRT] Tactic: 10354687 Time: 0.456927 [TRT] Tactic: 10551295 Time: 0.399505 [TRT] Tactic: 10747903 Time: 0.319115 [TRT] Tactic: 10944511 Time: 0.442162 [TRT] Fastest Tactic: 10747903 Time: 0.319115 [TRT] --------------- Timing Runner: inception_3b/pool_proj + inception_3b/relu_pool_proj (CudnnConvolution) [TRT] Tactic: 0 Time: 0.380572 [TRT] Tactic: 1 Time: 0.324791 [TRT] Tactic: 2 Time: 0.697136 [TRT] Tactic: 4 skipped. Scratch requested: 144900096, available: 33554432 [TRT] Tactic: 5 Time: 1.31724 [TRT] Fastest Tactic: 1 Time: 0.324791 [TRT] --------------- Timing Runner: inception_3b/pool_proj + inception_3b/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/pool_proj + inception_3b/relu_pool_proj (CaskConvolution) [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.24099 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v0 Tactic: 1698681053543049347 [TRT] Tactic: 1698681053543049347 Time: 0.227214 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.340026 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.180495 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.336771 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v1 Tactic: 5326823351883942011 [TRT] Tactic: 5326823351883942011 Time: 0.327708 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v0 Tactic: 5500448035057547314 [TRT] Tactic: 5500448035057547314 Time: 0.19237 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.182344 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.242291 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.342526 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v1 Tactic: -6576203419454146580 [TRT] Tactic: -6576203419454146580 Time: 0.214036 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.253881 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.225182 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.203932 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.200599 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.333072 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v0 Tactic: -377491875521947884 [TRT] Tactic: -377491875521947884 Time: 0.333932 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v1 Tactic: -37215280111360163 [TRT] Tactic: -37215280111360163 Time: 0.173386 [TRT] Fastest Tactic: -37215280111360163 Time: 0.173386 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -37215280111360163 [TRT] *************** Autotuning format combination: Float(200704,1,7168,256) -> Float(376320,1,13440,480) *************** [TRT] --------------- Timing Runner: inception_3b/pool_proj + inception_3b/relu_pool_proj (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/pool_proj + inception_3b/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/pool_proj + inception_3b/relu_pool_proj (CaskConvolution) [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 3886731678879822788 [TRT] Tactic: 3886731678879822788 Time: 0.183073 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 6629944304117643200 [TRT] Tactic: 6629944304117643200 Time: 0.326016 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.329948 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.183177 [TRT] Fastest Tactic: 3886731678879822788 Time: 0.183073 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 3886731678879822788 [TRT] *************** Autotuning format combination: Half(200704,784,28,1) -> Half(376320,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/pool_proj + inception_3b/relu_pool_proj (CudnnConvolution) [TRT] Tactic: 0 Time: 0.38625 [TRT] Tactic: 1 Time: 0.343516 [TRT] Tactic: 2 Time: 0.678594 [TRT] Tactic: 4 skipped. Scratch requested: 144900096, available: 33554432 [TRT] Tactic: 5 Time: 1.27182 [TRT] Fastest Tactic: 1 Time: 0.343516 [TRT] --------------- Timing Runner: inception_3b/pool_proj + inception_3b/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/pool_proj + inception_3b/relu_pool_proj (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 1 [TRT] *************** Autotuning format combination: Half(100352,784:2,28,1) -> Half(376320,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/pool_proj + inception_3b/relu_pool_proj (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] *************** Autotuning format combination: Half(100352,784:2,28,1) -> Half(188160,784:2,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/pool_proj + inception_3b/relu_pool_proj (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.206198 [TRT] Tactic: 655359 Time: 0.330938 [TRT] Tactic: 786431 Time: 0.276354 [TRT] Tactic: 851967 Time: 0.365443 [TRT] Tactic: 1179647 Time: 0.200469 [TRT] Tactic: 1310719 Time: 0.396146 [TRT] Tactic: 1376255 Time: 0.172083 [TRT] Tactic: 1441791 Time: 0.222577 [TRT] Tactic: 1507327 Time: 0.339661 [TRT] Tactic: 1638399 Time: 0.255938 [TRT] Tactic: 1835007 Time: 0.257734 [TRT] Tactic: 1900543 Time: 0.353255 [TRT] Tactic: 2097151 Time: 0.315573 [TRT] Tactic: 2162687 Time: 0.222318 [TRT] Tactic: 2293759 Time: 0.20625 [TRT] Tactic: 2359295 Time: 0.192865 [TRT] Tactic: 2686975 Time: 0.298776 [TRT] Tactic: 3080191 Time: 0.347552 [TRT] Tactic: 3342335 Time: 0.370676 [TRT] Tactic: 3407871 Time: 0.207265 [TRT] Tactic: 3538943 Time: 0.222943 [TRT] Tactic: 3670015 Time: 0.397839 [TRT] Tactic: 3932159 Time: 0.309974 [TRT] Tactic: 3997695 Time: 0.335677 [TRT] Tactic: 4063231 Time: 0.420079 [TRT] Tactic: 4194303 Time: 0.265156 [TRT] Tactic: 4259839 Time: 0.337656 [TRT] Tactic: 4325375 Time: 0.259479 [TRT] Tactic: 4521983 Time: 0.262161 [TRT] Tactic: 4587519 Time: 0.361224 [TRT] Tactic: 4653055 Time: 0.294453 [TRT] Tactic: 4915199 Time: 0.269791 [TRT] Tactic: 4980735 Time: 0.273229 [TRT] Tactic: 5177343 Time: 0.22 [TRT] Tactic: 5242879 Time: 0.201407 [TRT] Tactic: 5373951 Time: 0.299792 [TRT] Tactic: 5439487 Time: 0.359583 [TRT] Tactic: 5570559 Time: 0.37586 [TRT] Tactic: 5636095 Time: 0.418542 [TRT] Tactic: 5701631 Time: 0.203047 [TRT] Tactic: 5767167 Time: 0.645599 [TRT] Tactic: 5832703 Time: 0.213203 [TRT] Tactic: 5898239 Time: 0.224532 [TRT] Tactic: 6029311 Time: 0.208385 [TRT] Tactic: 6225919 Time: 0.195312 [TRT] Tactic: 6291455 Time: 0.234636 [TRT] Tactic: 6422527 Time: 0.338803 [TRT] Tactic: 6750207 Time: 0.28052 [TRT] Tactic: 6815743 Time: 0.241875 [TRT] Tactic: 6946815 Time: 0.370495 [TRT] Tactic: 7012351 Time: 0.345547 [TRT] Tactic: 7077887 Time: 0.209479 [TRT] Tactic: 7143423 Time: 0.400104 [TRT] Tactic: 7208959 Time: 0.211901 [TRT] Tactic: 7340031 Time: 0.231067 [TRT] Tactic: 7405567 Time: 0.230182 [TRT] Tactic: 7536639 Time: 0.250573 [TRT] Tactic: 7602175 Time: 0.293802 [TRT] Tactic: 7733247 Time: 0.208802 [TRT] Tactic: 7798783 Time: 0.327552 [TRT] Tactic: 8191999 Time: 0.397891 [TRT] Tactic: 8257535 Time: 0.268489 [TRT] Tactic: 8323071 Time: 0.281276 [TRT] Tactic: 8650751 Time: 0.289662 [TRT] Tactic: 8716287 Time: 0.250182 [TRT] Tactic: 9109503 Time: 0.341563 [TRT] Tactic: 9568255 Time: 0.26875 [TRT] Tactic: 9895935 Time: 0.291224 [TRT] Tactic: 10223615 Time: 0.347552 [TRT] Tactic: 10354687 Time: 0.367603 [TRT] Tactic: 10551295 Time: 0.263256 [TRT] Tactic: 10747903 Time: 0.214244 [TRT] Tactic: 10944511 Time: 0.303073 [TRT] Fastest Tactic: 1376255 Time: 0.172083 [TRT] --------------- Timing Runner: inception_3b/pool_proj + inception_3b/relu_pool_proj (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/pool_proj + inception_3b/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_3b/pool_proj + inception_3b/relu_pool_proj (CaskConvolution) [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_interior_nn_v1 Tactic: 3066127711859985668 [TRT] Tactic: 3066127711859985668 Time: 0.145625 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.16138 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.153672 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.13052 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] Tactic: 8163473458334948789 Time: 0.125781 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.231537 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.23638 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.128828 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_interior_nn_v1 Tactic: -1716393687483585322 [TRT] Tactic: -1716393687483585322 Time: 0.229844 [TRT] Fastest Tactic: 8163473458334948789 Time: 0.125781 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 8163473458334948789 [TRT] *************** Autotuning Reformat:Float(376320,784,28,1) -> Float(376320,1,13440,480) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.077474 [TRT] Tactic: 0 Time: 0.106119 [TRT] Fastest Tactic: 1002 Time: 0.077474 [TRT] *************** Autotuning Reformat:Float(376320,784,28,1) -> Half(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.13401 [TRT] Tactic: 0 Time: 0.0973955 [TRT] Fastest Tactic: 0 Time: 0.0973955 [TRT] *************** Autotuning Reformat:Float(376320,784,28,1) -> Half(188160,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.106432 [TRT] Tactic: 0 Time: 0.060182 [TRT] Fastest Tactic: 0 Time: 0.060182 [TRT] *************** Autotuning Reformat:Float(376320,1,13440,480) -> Float(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.100495 [TRT] Tactic: 0 Time: 0.103204 [TRT] Fastest Tactic: 1002 Time: 0.100495 [TRT] *************** Autotuning Reformat:Float(376320,1,13440,480) -> Half(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0701045 [TRT] Tactic: 0 Time: 0.0979945 [TRT] Fastest Tactic: 1002 Time: 0.0701045 [TRT] *************** Autotuning Reformat:Float(376320,1,13440,480) -> Half(188160,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0949485 [TRT] Tactic: 0 Time: 0.116563 [TRT] Fastest Tactic: 1002 Time: 0.0949485 [TRT] *************** Autotuning Reformat:Half(376320,784,28,1) -> Float(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.135911 [TRT] Tactic: 0 Time: 0.09875 [TRT] Fastest Tactic: 0 Time: 0.09875 [TRT] *************** Autotuning Reformat:Half(376320,784,28,1) -> Float(376320,1,13440,480) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.066875 [TRT] Tactic: 0 Time: 0.106119 [TRT] Fastest Tactic: 1002 Time: 0.066875 [TRT] *************** Autotuning Reformat:Half(376320,784,28,1) -> Half(188160,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0679425 [TRT] Tactic: 0 Time: 0.059505 [TRT] Fastest Tactic: 0 Time: 0.059505 [TRT] *************** Autotuning Reformat:Half(188160,784:2,28,1) -> Float(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0900785 [TRT] Tactic: 0 Time: 0.114167 [TRT] Fastest Tactic: 1002 Time: 0.0900785 [TRT] *************** Autotuning Reformat:Half(188160,784:2,28,1) -> Float(376320,1,13440,480) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0671355 [TRT] Tactic: 0 Time: 0.122057 [TRT] Fastest Tactic: 1002 Time: 0.0671355 [TRT] *************** Autotuning Reformat:Half(188160,784:2,28,1) -> Half(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.128854 [TRT] Tactic: 0 Time: 0.10263 [TRT] Fastest Tactic: 0 Time: 0.10263 [TRT] *************** Autotuning Reformat:Float(225792,784,28,1) -> Float(376320,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.26138 [TRT] Tactic: 0 Time: 0.0611715 [TRT] Fastest Tactic: 0 Time: 0.0611715 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(225792,784,28,1) -> Float(376320,1,13440,480) *************** [TRT] --------------- Timing Runner: inception_3b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.141511 [TRT] Tactic: 0 Time: 0.204167 [TRT] Fastest Tactic: 1002 Time: 0.141511 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(225792,784,28,1) -> Half(376320,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.259791 [TRT] Tactic: 0 Time: 0.188515 [TRT] Fastest Tactic: 0 Time: 0.188515 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(225792,784,28,1) -> Half(188160,784:2,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.201614 [TRT] Tactic: 0 Time: 0.224661 [TRT] Fastest Tactic: 1002 Time: 0.201614 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(225792,1,8064,288) -> Float(376320,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.175624 [TRT] Tactic: 0 Time: 0.208229 [TRT] Fastest Tactic: 1002 Time: 0.175624 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(225792,1,8064,288) -> Float(376320,1,13440,480) *************** [TRT] --------------- Timing Runner: inception_3b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.289193 [TRT] Tactic: 0 Time: 0.201329 [TRT] Fastest Tactic: 0 Time: 0.201329 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(225792,1,8064,288) -> Half(376320,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.134193 [TRT] Tactic: 0 Time: 0.206328 [TRT] Fastest Tactic: 1002 Time: 0.134193 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(225792,1,8064,288) -> Half(188160,784:2,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.178359 [TRT] Tactic: 0 Time: 0.236562 [TRT] Fastest Tactic: 1002 Time: 0.178359 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(225792,784,28,1) -> Float(376320,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.262604 [TRT] Tactic: 0 Time: 0.189479 [TRT] Fastest Tactic: 0 Time: 0.189479 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(225792,784,28,1) -> Float(376320,1,13440,480) *************** [TRT] --------------- Timing Runner: inception_3b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.127109 [TRT] Tactic: 0 Time: 0.206198 [TRT] Fastest Tactic: 1002 Time: 0.127109 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(225792,784,28,1) -> Half(376320,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.179375 [TRT] Tactic: 0 Time: 0.032474 [TRT] Fastest Tactic: 0 Time: 0.032474 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(225792,784,28,1) -> Half(188160,784:2,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.13362 [TRT] Tactic: 0 Time: 0.112578 [TRT] Fastest Tactic: 0 Time: 0.112578 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(112896,784:2,28,1) -> Float(376320,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.169088 [TRT] Tactic: 0 Time: 0.221146 [TRT] Fastest Tactic: 1002 Time: 0.169088 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(112896,784:2,28,1) -> Float(376320,1,13440,480) *************** [TRT] --------------- Timing Runner: inception_3b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.128906 [TRT] Tactic: 0 Time: 0.238854 [TRT] Fastest Tactic: 1002 Time: 0.128906 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(112896,784:2,28,1) -> Half(376320,784,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.239844 [TRT] Tactic: 0 Time: 0.200833 [TRT] Fastest Tactic: 0 Time: 0.200833 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(112896,784:2,28,1) -> Half(188160,784:2,28,1) *************** [TRT] --------------- Timing Runner: inception_3b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.206484 [TRT] Tactic: 0 Time: 0.0331245 [TRT] Fastest Tactic: 0 Time: 0.0331245 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(376320,784,28,1) -> Half(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.806067 [TRT] Tactic: 0 Time: 0.512344 [TRT] Fastest Tactic: 0 Time: 0.512344 [TRT] *************** Autotuning Reformat:Float(376320,784,28,1) -> Half(188160,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.728412 [TRT] Tactic: 0 Time: 0.411797 [TRT] Fastest Tactic: 0 Time: 0.411797 [TRT] *************** Autotuning Reformat:Float(376320,1,13440,480) -> Float(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.642031 [TRT] Tactic: 0 Time: 0.763985 [TRT] Fastest Tactic: 1002 Time: 0.642031 [TRT] *************** Autotuning Reformat:Float(376320,1,13440,480) -> Half(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.481953 [TRT] Tactic: 0 Time: 0.756328 [TRT] Fastest Tactic: 1002 Time: 0.481953 [TRT] *************** Autotuning Reformat:Float(376320,1,13440,480) -> Half(188160,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.637031 [TRT] Tactic: 0 Time: 0.863359 [TRT] Fastest Tactic: 1002 Time: 0.637031 [TRT] *************** Autotuning Reformat:Half(376320,784,28,1) -> Float(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.815026 [TRT] Tactic: 0 Time: 0.43625 [TRT] Fastest Tactic: 0 Time: 0.43625 [TRT] *************** Autotuning Reformat:Half(376320,784,28,1) -> Half(188160,784:2,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.459453 [TRT] Tactic: 0 Time: 0.40664 [TRT] Fastest Tactic: 0 Time: 0.40664 [TRT] *************** Autotuning Reformat:Half(188160,784:2,28,1) -> Float(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.597813 [TRT] Tactic: 0 Time: 0.351928 [TRT] Fastest Tactic: 0 Time: 0.351928 [TRT] *************** Autotuning Reformat:Half(188160,784:2,28,1) -> Half(376320,784,28,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.866354 [TRT] Tactic: 0 Time: 0.345625 [TRT] Fastest Tactic: 0 Time: 0.345625 [TRT] *************** Autotuning format combination: Float(376320,784,28,1) -> Float(94080,196,14,1) *************** [TRT] --------------- Timing Runner: pool3/3x3_s2 (TiledPooling) [TRT] Tactic: 257 Time: 0.58263 [TRT] Tactic: 65793 Time: 0.563984 [TRT] Tactic: 131329 Time: 0.676615 [TRT] Tactic: 196865 Time: 2.89693 [TRT] Tactic: 262401 Time: 2.65497 [TRT] Tactic: 327937 Time: 1.36081 [TRT] Tactic: 393473 Time: 1.29495 [TRT] Tactic: 459009 Time: 0.347969 [TRT] Tactic: 524545 Time: 0.339844 [TRT] Tactic: 590081 Time: 0.417448 [TRT] Tactic: 655617 Time: 1.7249 [TRT] Tactic: 721153 Time: 1.57461 [TRT] Tactic: 786689 Time: 0.815156 [TRT] Tactic: 852225 Time: 0.757786 [TRT] Tactic: 917761 Time: 0.276458 [TRT] Tactic: 983297 Time: 0.272839 [TRT] Tactic: 1048833 Time: 0.315573 [TRT] Tactic: 1114369 Time: 1.28974 [TRT] Tactic: 1179905 Time: 1.2212 [TRT] Tactic: 1245441 Time: 0.630338 [TRT] Tactic: 1310977 Time: 0.586093 [TRT] Tactic: 1376513 Time: 0.231406 [TRT] Tactic: 1442049 Time: 0.227136 [TRT] Tactic: 1507585 Time: 0.265468 [TRT] Tactic: 1573121 Time: 1.07031 [TRT] Tactic: 1638657 Time: 1.01997 [TRT] Tactic: 1704193 Time: 0.478151 [TRT] Tactic: 1769729 Time: 0.467891 [TRT] Tactic: 1835265 Time: 0.213828 [TRT] Tactic: 1900801 Time: 0.206979 [TRT] Tactic: 1966337 Time: 0.246432 [TRT] Tactic: 2031873 Time: 0.952682 [TRT] Tactic: 2097409 Time: 0.926171 [TRT] Tactic: 2162945 Time: 0.42289 [TRT] Tactic: 2228481 Time: 0.432083 [TRT] Tactic: 2294017 Time: 0.212603 [TRT] Tactic: 2359553 Time: 0.202708 [TRT] Tactic: 2425089 Time: 0.222813 [TRT] Tactic: 2490625 Time: 0.869583 [TRT] Tactic: 2556161 Time: 0.831927 [TRT] Tactic: 2621697 Time: 0.379896 [TRT] Tactic: 2687233 Time: 0.392474 [TRT] Tactic: 6947073 Time: 0.468672 [TRT] Fastest Tactic: 2359553 Time: 0.202708 [TRT] --------------- Timing Runner: pool3/3x3_s2 (CudnnPooling) [TRT] Tactic: -1 Time: 0.86323 [TRT] Fastest Tactic: -1 Time: 0.86323 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: TiledPooling Tactic: 2359553 [TRT] *************** Autotuning format combination: Half(376320,784,28,1) -> Half(94080,196,14,1) *************** [TRT] --------------- Timing Runner: pool3/3x3_s2 (TiledPooling) [TRT] TiledPooling has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: pool3/3x3_s2 (CudnnPooling) [TRT] Tactic: -1 Time: 0.904688 [TRT] Fastest Tactic: -1 Time: 0.904688 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnPooling Tactic: -1 [TRT] *************** Autotuning format combination: Half(188160,784:2,28,1) -> Half(47040,196:2,14,1) *************** [TRT] --------------- Timing Runner: pool3/3x3_s2 (TiledPooling) [TRT] Tactic: 257 Time: 0.306823 [TRT] Tactic: 65793 Time: 0.295885 [TRT] Tactic: 131329 Time: 0.353776 [TRT] Tactic: 196865 Time: 1.56013 [TRT] Tactic: 262401 Time: 1.4106 [TRT] Tactic: 327937 Time: 0.742136 [TRT] Tactic: 393473 Time: 0.686978 [TRT] Tactic: 459009 Time: 0.199557 [TRT] Tactic: 524545 Time: 0.191094 [TRT] Tactic: 590081 Time: 0.226041 [TRT] Tactic: 655617 Time: 0.94776 [TRT] Tactic: 721153 Time: 0.90237 [TRT] Tactic: 786689 Time: 0.439453 [TRT] Tactic: 852225 Time: 0.422839 [TRT] Tactic: 917761 Time: 0.153854 [TRT] Tactic: 983297 Time: 0.15013 [TRT] Tactic: 1048833 Time: 0.172969 [TRT] Tactic: 1114369 Time: 0.747656 [TRT] Tactic: 1179905 Time: 0.693594 [TRT] Tactic: 1245441 Time: 0.326276 [TRT] Tactic: 1310977 Time: 0.33 [TRT] Tactic: 1376513 Time: 0.133255 [TRT] Tactic: 1442049 Time: 0.133594 [TRT] Tactic: 1507585 Time: 0.15086 [TRT] Tactic: 1573121 Time: 0.622005 [TRT] Tactic: 1638657 Time: 0.591823 [TRT] Tactic: 1704193 Time: 0.271719 [TRT] Tactic: 1769729 Time: 0.27974 [TRT] Tactic: 1835265 Time: 0.124011 [TRT] Tactic: 1900801 Time: 0.124011 [TRT] Tactic: 1966337 Time: 0.140443 [TRT] Tactic: 2031873 Time: 0.605286 [TRT] Tactic: 2097409 Time: 0.555573 [TRT] Tactic: 2162945 Time: 0.251693 [TRT] Tactic: 2228481 Time: 0.270469 [TRT] Tactic: 2294017 Time: 0.117708 [TRT] Tactic: 2359553 Time: 0.117136 [TRT] Tactic: 2425089 Time: 0.132161 [TRT] Tactic: 2490625 Time: 0.50888 [TRT] Tactic: 2556161 Time: 0.510078 [TRT] Tactic: 2621697 Time: 0.23125 [TRT] Tactic: 2687233 Time: 0.23776 [TRT] Tactic: 6947073 Time: 0.279635 [TRT] Fastest Tactic: 2359553 Time: 0.117136 [TRT] --------------- Timing Runner: pool3/3x3_s2 (CudaPooling) [TRT] Tactic: -3 Time: 0.227266 [TRT] Fastest Tactic: -3 Time: 0.227266 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: TiledPooling Tactic: 2359553 [TRT] *************** Autotuning Reformat:Float(94080,196,14,1) -> Float(94080,1,6720,480) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.173568 [TRT] Tactic: 0 Time: 0.190729 [TRT] Fastest Tactic: 1002 Time: 0.173568 [TRT] *************** Autotuning Reformat:Float(94080,196,14,1) -> Half(94080,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 3.38669 [TRT] Tactic: 0 Time: 0.134219 [TRT] Fastest Tactic: 0 Time: 0.134219 [TRT] *************** Autotuning Reformat:Float(94080,196,14,1) -> Half(47040,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.204401 [TRT] Tactic: 0 Time: 0.107552 [TRT] Fastest Tactic: 0 Time: 0.107552 [TRT] *************** Autotuning Reformat:Half(94080,196,14,1) -> Float(94080,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 3.4906 [TRT] Tactic: 0 Time: 0.113333 [TRT] Fastest Tactic: 0 Time: 0.113333 [TRT] *************** Autotuning Reformat:Half(94080,196,14,1) -> Float(94080,1,6720,480) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.134141 [TRT] Tactic: 0 Time: 0.193177 [TRT] Fastest Tactic: 1002 Time: 0.134141 [TRT] *************** Autotuning Reformat:Half(94080,196,14,1) -> Half(47040,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.145781 [TRT] Tactic: 0 Time: 0.106146 [TRT] Fastest Tactic: 0 Time: 0.106146 [TRT] *************** Autotuning Reformat:Half(47040,196:2,14,1) -> Float(94080,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.189557 [TRT] Tactic: 0 Time: 0.092942 [TRT] Fastest Tactic: 0 Time: 0.092942 [TRT] *************** Autotuning Reformat:Half(47040,196:2,14,1) -> Float(94080,1,6720,480) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.134453 [TRT] Tactic: 0 Time: 0.224714 [TRT] Fastest Tactic: 1002 Time: 0.134453 [TRT] *************** Autotuning Reformat:Half(47040,196:2,14,1) -> Half(94080,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.26901 [TRT] Tactic: 0 Time: 0.091432 [TRT] Fastest Tactic: 0 Time: 0.091432 [TRT] *************** Autotuning Reformat:Float(94080,196,14,1) -> Float(94080,1,6720,480) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.173984 [TRT] Tactic: 0 Time: 0.19112 [TRT] Fastest Tactic: 1002 Time: 0.173984 [TRT] *************** Autotuning Reformat:Float(94080,196,14,1) -> Half(94080,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 3.38641 [TRT] Tactic: 0 Time: 0.134167 [TRT] Fastest Tactic: 0 Time: 0.134167 [TRT] *************** Autotuning Reformat:Float(94080,196,14,1) -> Half(47040,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.204401 [TRT] Tactic: 0 Time: 0.107369 [TRT] Fastest Tactic: 0 Time: 0.107369 [TRT] *************** Autotuning Reformat:Float(94080,1,6720,480) -> Float(94080,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.216719 [TRT] Tactic: 0 Time: 0.176693 [TRT] Fastest Tactic: 0 Time: 0.176693 [TRT] *************** Autotuning Reformat:Float(94080,1,6720,480) -> Half(94080,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.145364 [TRT] Tactic: 0 Time: 0.178541 [TRT] Fastest Tactic: 1002 Time: 0.145364 [TRT] *************** Autotuning Reformat:Float(94080,1,6720,480) -> Half(47040,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.177187 [TRT] Tactic: 0 Time: 0.212344 [TRT] Fastest Tactic: 1002 Time: 0.177187 [TRT] *************** Autotuning Reformat:Half(94080,196,14,1) -> Float(94080,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 3.49349 [TRT] Tactic: 0 Time: 0.102057 [TRT] Fastest Tactic: 0 Time: 0.102057 [TRT] *************** Autotuning Reformat:Half(94080,196,14,1) -> Float(94080,1,6720,480) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.122031 [TRT] Tactic: 0 Time: 0.175287 [TRT] Fastest Tactic: 1002 Time: 0.122031 [TRT] *************** Autotuning Reformat:Half(94080,196,14,1) -> Half(47040,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.134297 [TRT] Tactic: 0 Time: 0.096407 [TRT] Fastest Tactic: 0 Time: 0.096407 [TRT] *************** Autotuning Reformat:Half(47040,196:2,14,1) -> Float(94080,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.171406 [TRT] Tactic: 0 Time: 0.0847915 [TRT] Fastest Tactic: 0 Time: 0.0847915 [TRT] *************** Autotuning Reformat:Half(47040,196:2,14,1) -> Float(94080,1,6720,480) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.121954 [TRT] Tactic: 0 Time: 0.202395 [TRT] Fastest Tactic: 1002 Time: 0.121954 [TRT] *************** Autotuning Reformat:Half(47040,196:2,14,1) -> Half(94080,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.240625 [TRT] Tactic: 0 Time: 0.082917 [TRT] Fastest Tactic: 0 Time: 0.082917 [TRT] *************** Autotuning format combination: Float(94080,196,14,1) -> Float(59584,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 1.00208 [TRT] Tactic: 655359 Time: 0.846328 [TRT] Tactic: 786431 Time: 1.05698 [TRT] Tactic: 851967 Time: 1.2799 [TRT] Tactic: 1179647 Time: 1.14393 [TRT] Tactic: 1310719 Time: 2.25865 [TRT] Tactic: 1376255 Time: 0.839947 [TRT] Tactic: 1441791 Time: 1.33667 [TRT] Tactic: 1507327 Time: 1.31495 [TRT] Tactic: 1638399 Time: 1.55932 [TRT] Tactic: 1835007 Time: 1.13609 [TRT] Tactic: 1900543 Time: 1.24534 [TRT] Tactic: 2097151 Time: 1.21281 [TRT] Tactic: 2162687 Time: 0.862943 [TRT] Tactic: 2293759 Time: 0.89737 [TRT] Tactic: 2359295 Time: 1.06758 [TRT] Tactic: 2686975 Time: 0.987292 [TRT] Tactic: 3080191 Time: 1.00432 [TRT] Tactic: 3342335 Time: 1.22128 [TRT] Tactic: 3407871 Time: 0.934532 [TRT] Tactic: 3538943 Time: 0.967864 [TRT] Tactic: 3670015 Time: 0.784922 [TRT] Tactic: 3932159 Time: 1.36448 [TRT] Tactic: 3997695 Time: 1.07211 [TRT] Tactic: 4063231 Time: 1.115 [TRT] Tactic: 4194303 Time: 1.03839 [TRT] Tactic: 4259839 Time: 1.2749 [TRT] Tactic: 4325375 Time: 1.31177 [TRT] Tactic: 4521983 Time: 1.30526 [TRT] Tactic: 4587519 Time: 1.21057 [TRT] Tactic: 4653055 Time: 1.14763 [TRT] Tactic: 4915199 Time: 1.03531 [TRT] Tactic: 4980735 Time: 1.39503 [TRT] Tactic: 5177343 Time: 1.35667 [TRT] Tactic: 5242879 Time: 0.838698 [TRT] Tactic: 5373951 Time: 1.27096 [TRT] Tactic: 5439487 Time: 1.10724 [TRT] Tactic: 5570559 Time: 0.896744 [TRT] Tactic: 5636095 Time: 1.1231 [TRT] Tactic: 5701631 Time: 1.08143 [TRT] Tactic: 5767167 Time: 1.65651 [TRT] Tactic: 5832703 Time: 0.933203 [TRT] Tactic: 5898239 Time: 0.824348 [TRT] Tactic: 6029311 Time: 0.840781 [TRT] Tactic: 6225919 Time: 0.898541 [TRT] Tactic: 6291455 Time: 1.13154 [TRT] Tactic: 6422527 Time: 0.99125 [TRT] Tactic: 6750207 Time: 1.01443 [TRT] Tactic: 6815743 Time: 0.912084 [TRT] Tactic: 6946815 Time: 1.37982 [TRT] Tactic: 7012351 Time: 1.21294 [TRT] Tactic: 7077887 Time: 0.97487 [TRT] Tactic: 7143423 Time: 1.61404 [TRT] Tactic: 7208959 Time: 1.09448 [TRT] Tactic: 7340031 Time: 0.886589 [TRT] Tactic: 7405567 Time: 1.01273 [TRT] Tactic: 7536639 Time: 1.03341 [TRT] Tactic: 7602175 Time: 1.29513 [TRT] Tactic: 7733247 Time: 0.922292 [TRT] Tactic: 7798783 Time: 1.05641 [TRT] Tactic: 8191999 Time: 1.64937 [TRT] Tactic: 8257535 Time: 1.05565 [TRT] Tactic: 8323071 Time: 0.97763 [TRT] Tactic: 8650751 Time: 1.51833 [TRT] Tactic: 8716287 Time: 1.00089 [TRT] Tactic: 9109503 Time: 1.28307 [TRT] Tactic: 9568255 Time: 1.02612 [TRT] Tactic: 9895935 Time: 1.03797 [TRT] Tactic: 10223615 Time: 0.986276 [TRT] Tactic: 10354687 Time: 1.28802 [TRT] Tactic: 10551295 Time: 0.957995 [TRT] Tactic: 10747903 Time: 0.868307 [TRT] Tactic: 10944511 Time: 1.39854 [TRT] Fastest Tactic: 3670015 Time: 0.784922 [TRT] --------------- Timing Runner: inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce (CudnnConvolution) [TRT] Tactic: 0 Time: 1.07005 [TRT] Tactic: 1 Time: 0.900885 [TRT] Tactic: 2 Time: 1.1825 [TRT] Tactic: 4 skipped. Scratch requested: 337889280, available: 33554432 [TRT] Tactic: 5 Time: 12.3602 [TRT] Fastest Tactic: 1 Time: 0.900885 [TRT] --------------- Timing Runner: inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce (CaskConvolution) [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.592031 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v0 Tactic: 1698681053543049347 [TRT] Tactic: 1698681053543049347 Time: 0.565678 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.555261 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.46638 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.549453 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v1 Tactic: 5326823351883942011 [TRT] Tactic: 5326823351883942011 Time: 0.530364 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v0 Tactic: 5500448035057547314 [TRT] Tactic: 5500448035057547314 Time: 0.534844 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.48138 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.613724 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.558594 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v1 Tactic: -6576203419454146580 [TRT] Tactic: -6576203419454146580 Time: 0.516719 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.624688 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.543958 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.544192 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.539011 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.542604 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v0 Tactic: -377491875521947884 [TRT] Tactic: -377491875521947884 Time: 0.543021 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v1 Tactic: -37215280111360163 [TRT] Tactic: -37215280111360163 Time: 0.458542 [TRT] Fastest Tactic: -37215280111360163 Time: 0.458542 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -37215280111360163 [TRT] *************** Autotuning format combination: Float(94080,1,6720,480) -> Float(59584,1,4256,304) *************** [TRT] --------------- Timing Runner: inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce (CaskConvolution) [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 3886731678879822788 [TRT] Tactic: 3886731678879822788 Time: 0.467447 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 6629944304117643200 [TRT] Tactic: 6629944304117643200 Time: 0.768958 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.77711 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.463359 [TRT] Fastest Tactic: -7394439838318485025 Time: 0.463359 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(94080,196,14,1) -> Half(59584,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce (CudnnConvolution) [TRT] Tactic: 0 Time: 1.06562 [TRT] Tactic: 1 Time: 0.847656 [TRT] Tactic: 2 Time: 1.05089 [TRT] Tactic: 4 skipped. Scratch requested: 337889280, available: 33554432 [TRT] Tactic: 5 Time: 10.4665 [TRT] Fastest Tactic: 1 Time: 0.847656 [TRT] --------------- Timing Runner: inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 1 [TRT] *************** Autotuning format combination: Half(47040,196:2,14,1) -> Half(59584,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] *************** Autotuning format combination: Half(47040,196:2,14,1) -> Half(29792,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.449791 [TRT] Tactic: 655359 Time: 0.474193 [TRT] Tactic: 786431 Time: 0.592682 [TRT] Tactic: 851967 Time: 0.632032 [TRT] Tactic: 1179647 Time: 0.442526 [TRT] Tactic: 1310719 Time: 0.936016 [TRT] Tactic: 1376255 Time: 0.356328 [TRT] Tactic: 1441791 Time: 0.591277 [TRT] Tactic: 1507327 Time: 0.643776 [TRT] Tactic: 1638399 Time: 0.731822 [TRT] Tactic: 1835007 Time: 0.611875 [TRT] Tactic: 1900543 Time: 0.596068 [TRT] Tactic: 2162687 Time: 0.375989 [TRT] Tactic: 2293759 Time: 0.402786 [TRT] Tactic: 2359295 Time: 0.477421 [TRT] Tactic: 2686975 Time: 0.665807 [TRT] Tactic: 3080191 Time: 0.479088 [TRT] Tactic: 3342335 Time: 0.596042 [TRT] Tactic: 3407871 Time: 0.406172 [TRT] Tactic: 3538943 Time: 0.433359 [TRT] Tactic: 3670015 Time: 0.452995 [TRT] Tactic: 3932159 Time: 0.553411 [TRT] Tactic: 3997695 Time: 0.587708 [TRT] Tactic: 4063231 Time: 0.534011 [TRT] Tactic: 4194303 Time: 0.511953 [TRT] Tactic: 4325375 Time: 0.58901 [TRT] Tactic: 4521983 Time: 0.596537 [TRT] Tactic: 4587519 Time: 0.614115 [TRT] Tactic: 4653055 Time: 0.527135 [TRT] Tactic: 4915199 Time: 0.51138 [TRT] Tactic: 4980735 Time: 0.596563 [TRT] Tactic: 5177343 Time: 0.530625 [TRT] Tactic: 5242879 Time: 0.353437 [TRT] Tactic: 5373951 Time: 0.522005 [TRT] Tactic: 5439487 Time: 0.537995 [TRT] Tactic: 5570559 Time: 0.522187 [TRT] Tactic: 5636095 Time: 0.538021 [TRT] Tactic: 5701631 Time: 0.42513 [TRT] Tactic: 5767167 Time: 0.66487 [TRT] Tactic: 5832703 Time: 0.393308 [TRT] Tactic: 5898239 Time: 0.442943 [TRT] Tactic: 6029311 Time: 0.355781 [TRT] Tactic: 6225919 Time: 0.383177 [TRT] Tactic: 6291455 Time: 0.444297 [TRT] Tactic: 6422527 Time: 0.420287 [TRT] Tactic: 6750207 Time: 0.515625 [TRT] Tactic: 6815743 Time: 0.397109 [TRT] Tactic: 6946815 Time: 0.599244 [TRT] Tactic: 7077887 Time: 0.41513 [TRT] Tactic: 7143423 Time: 0.688697 [TRT] Tactic: 7208959 Time: 0.460442 [TRT] Tactic: 7340031 Time: 0.473099 [TRT] Tactic: 7405567 Time: 0.451771 [TRT] Tactic: 7536639 Time: 0.468828 [TRT] Tactic: 7602175 Time: 0.546328 [TRT] Tactic: 7733247 Time: 0.445937 [TRT] Tactic: 7798783 Time: 0.591302 [TRT] Tactic: 8191999 Time: 0.717553 [TRT] Tactic: 8257535 Time: 0.525781 [TRT] Tactic: 8323071 Time: 0.487005 [TRT] Tactic: 8650751 Time: 0.644166 [TRT] Tactic: 8716287 Time: 0.437838 [TRT] Tactic: 9568255 Time: 0.51388 [TRT] Tactic: 9895935 Time: 0.512682 [TRT] Tactic: 10223615 Time: 0.6681 [TRT] Tactic: 10354687 Time: 0.725183 [TRT] Tactic: 10551295 Time: 0.410079 [TRT] Tactic: 10747903 Time: 0.391744 [TRT] Tactic: 10944511 Time: 0.59789 [TRT] Fastest Tactic: 5242879 Time: 0.353437 [TRT] --------------- Timing Runner: inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce (CaskConvolution) [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_interior_nn_v1 Tactic: 3066127711859985668 [TRT] Tactic: 3066127711859985668 Time: 0.247317 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.291484 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.257031 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.233204 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] Tactic: 8163473458334948789 Time: 0.218801 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.259297 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.264453 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.227422 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_interior_nn_v1 Tactic: -1716393687483585322 [TRT] Tactic: -1716393687483585322 Time: 0.25638 [TRT] Fastest Tactic: 8163473458334948789 Time: 0.218801 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 8163473458334948789 [TRT] *************** Autotuning Reformat:Float(59584,196,14,1) -> Float(59584,1,4256,304) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0887765 [TRT] Tactic: 0 Time: 0.096875 [TRT] Fastest Tactic: 1002 Time: 0.0887765 [TRT] *************** Autotuning Reformat:Float(59584,196,14,1) -> Half(59584,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.60549 [TRT] Tactic: 0 Time: 0.068308 [TRT] Fastest Tactic: 0 Time: 0.068308 [TRT] *************** Autotuning Reformat:Float(59584,196,14,1) -> Half(29792,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.101016 [TRT] Tactic: 0 Time: 0.0533595 [TRT] Fastest Tactic: 0 Time: 0.0533595 [TRT] *************** Autotuning Reformat:Float(59584,1,4256,304) -> Float(59584,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.108776 [TRT] Tactic: 0 Time: 0.0876565 [TRT] Fastest Tactic: 0 Time: 0.0876565 [TRT] *************** Autotuning Reformat:Float(59584,1,4256,304) -> Half(59584,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.075677 [TRT] Tactic: 0 Time: 0.088412 [TRT] Fastest Tactic: 1002 Time: 0.075677 [TRT] *************** Autotuning Reformat:Float(59584,1,4256,304) -> Half(29792,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0891925 [TRT] Tactic: 0 Time: 0.103959 [TRT] Fastest Tactic: 1002 Time: 0.0891925 [TRT] *************** Autotuning Reformat:Half(59584,196,14,1) -> Float(59584,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.65487 [TRT] Tactic: 0 Time: 0.056172 [TRT] Fastest Tactic: 0 Time: 0.056172 [TRT] *************** Autotuning Reformat:Half(59584,196,14,1) -> Float(59584,1,4256,304) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.069739 [TRT] Tactic: 0 Time: 0.0947135 [TRT] Fastest Tactic: 1002 Time: 0.069739 [TRT] *************** Autotuning Reformat:Half(59584,196,14,1) -> Half(29792,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0716925 [TRT] Tactic: 0 Time: 0.0526565 [TRT] Fastest Tactic: 0 Time: 0.0526565 [TRT] *************** Autotuning Reformat:Half(29792,196:2,14,1) -> Float(59584,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.095833 [TRT] Tactic: 0 Time: 0.046406 [TRT] Fastest Tactic: 0 Time: 0.046406 [TRT] *************** Autotuning Reformat:Half(29792,196:2,14,1) -> Float(59584,1,4256,304) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0712765 [TRT] Tactic: 0 Time: 0.109037 [TRT] Fastest Tactic: 1002 Time: 0.0712765 [TRT] *************** Autotuning Reformat:Half(29792,196:2,14,1) -> Half(59584,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.13112 [TRT] Tactic: 0 Time: 0.0444535 [TRT] Fastest Tactic: 0 Time: 0.0444535 [TRT] *************** Autotuning Reformat:Float(59584,196,14,1) -> Float(59584,1,4256,304) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0359115 [TRT] Tactic: 0 Time: 0.0324225 [TRT] Fastest Tactic: 0 Time: 0.0324225 [TRT] *************** Autotuning Reformat:Float(59584,196,14,1) -> Half(59584,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.616953 [TRT] Tactic: 0 Time: 0.030104 [TRT] Fastest Tactic: 0 Time: 0.030104 [TRT] *************** Autotuning Reformat:Float(59584,196,14,1) -> Half(29792,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0357815 [TRT] Tactic: 0 Time: 0.035599 [TRT] Fastest Tactic: 0 Time: 0.035599 [TRT] *************** Autotuning Reformat:Float(59584,1,4256,304) -> Float(59584,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.045026 [TRT] Tactic: 0 Time: 0.0289845 [TRT] Fastest Tactic: 0 Time: 0.0289845 [TRT] *************** Autotuning Reformat:Float(59584,1,4256,304) -> Half(59584,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.026745 [TRT] Tactic: 0 Time: 0.0301305 [TRT] Fastest Tactic: 1002 Time: 0.026745 [TRT] *************** Autotuning Reformat:Float(59584,1,4256,304) -> Half(29792,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.031198 [TRT] Tactic: 0 Time: 0.035781 [TRT] Fastest Tactic: 1002 Time: 0.031198 [TRT] *************** Autotuning Reformat:Half(59584,196,14,1) -> Float(59584,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.630989 [TRT] Tactic: 0 Time: 0.019479 [TRT] Fastest Tactic: 0 Time: 0.019479 [TRT] *************** Autotuning Reformat:Half(59584,196,14,1) -> Float(59584,1,4256,304) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0241925 [TRT] Tactic: 0 Time: 0.0320055 [TRT] Fastest Tactic: 1002 Time: 0.0241925 [TRT] *************** Autotuning Reformat:Half(59584,196,14,1) -> Half(29792,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.027891 [TRT] Tactic: 0 Time: 0.019479 [TRT] Fastest Tactic: 0 Time: 0.019479 [TRT] *************** Autotuning Reformat:Half(29792,196:2,14,1) -> Float(59584,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.035729 [TRT] Tactic: 0 Time: 0.0171615 [TRT] Fastest Tactic: 0 Time: 0.0171615 [TRT] *************** Autotuning Reformat:Half(29792,196:2,14,1) -> Float(59584,1,4256,304) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.024193 [TRT] Tactic: 0 Time: 0.037995 [TRT] Fastest Tactic: 1002 Time: 0.024193 [TRT] *************** Autotuning Reformat:Half(29792,196:2,14,1) -> Half(59584,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.044922 [TRT] Tactic: 0 Time: 0.031042 [TRT] Fastest Tactic: 0 Time: 0.031042 [TRT] *************** Autotuning format combination: Float(59584,196,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/3x3 + inception_4a/relu_3x3 (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4a/3x3 + inception_4a/relu_3x3 (FusedConvActConvolution) [TRT] Tactic: 524287 Time: 0.560104 [TRT] Tactic: 720895 Time: 0.68823 [TRT] Tactic: 983039 Time: 0.588542 [TRT] Tactic: 1048575 Time: 0.567188 [TRT] Tactic: 1703935 Time: 0.557214 [TRT] Tactic: 1769471 Time: 0.61474 [TRT] Tactic: 1966079 Time: 0.656433 [TRT] Tactic: 2031615 Time: 0.621094 [TRT] Tactic: 2228223 Time: 0.708984 [TRT] Tactic: 2424831 Time: 0.794609 [TRT] Tactic: 2621439 Time: 0.677266 [TRT] Tactic: 2752511 Time: 0.561954 [TRT] Tactic: 2818047 Time: 0.784765 [TRT] Tactic: 2883583 Time: 0.628906 [TRT] Tactic: 3014655 Time: 0.526823 [TRT] Tactic: 3145727 Time: 0.561927 [TRT] Tactic: 3473407 Time: 0.62638 [TRT] Tactic: 3604479 Time: 0.521875 [TRT] Tactic: 3735551 Time: 0.683177 [TRT] Tactic: 4390911 Time: 0.591614 [TRT] Tactic: 5046271 Time: 0.546589 [TRT] Tactic: 5963775 Time: 0.597734 [TRT] Tactic: 6160383 Time: 0.544662 [TRT] Tactic: 6488063 Time: 0.490182 [TRT] Tactic: 6881279 Time: 0.5719 [TRT] Tactic: 7274495 Time: 0.775261 [TRT] Tactic: 7864319 Time: 0.707109 [TRT] Tactic: 7995391 Time: 0.66625 [TRT] Tactic: 8585215 Time: 0.528463 [TRT] Tactic: 8847359 Time: 0.536041 [TRT] Tactic: 8978431 Time: 0.606979 [TRT] Tactic: 9043967 Time: 0.482083 [TRT] Tactic: 9175039 Time: 0.51862 [TRT] Tactic: 9502719 Time: 0.594532 [TRT] Tactic: 9830399 Time: 0.535807 [TRT] Tactic: 9961471 Time: 0.587865 [TRT] Tactic: 10027007 Time: 0.494922 [TRT] Tactic: 10092543 Time: 0.587214 [TRT] Tactic: 10289151 Time: 0.648177 [TRT] Tactic: 10485759 Time: 0.492422 [TRT] Tactic: 10682367 Time: 0.676276 [TRT] Tactic: 10813439 Time: 0.662135 [TRT] Fastest Tactic: 9043967 Time: 0.482083 [TRT] --------------- Timing Runner: inception_4a/3x3 + inception_4a/relu_3x3 (CudnnConvolution) [TRT] Tactic: 0 Time: 1.13383 [TRT] Tactic: 1 Time: 0.700781 [TRT] Tactic: 2 Time: 1.03357 [TRT] Tactic: 4 skipped. Scratch requested: 46946304, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 88223744, available: 33554432 [TRT] Tactic: 6 Time: 0.520391 [TRT] Fastest Tactic: 6 Time: 0.520391 [TRT] --------------- Timing Runner: inception_4a/3x3 + inception_4a/relu_3x3 (CaskConvolution) [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.723854 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v0 Tactic: 1754984623894446479 [TRT] Tactic: 1754984623894446479 Time: 0.769193 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v0 Tactic: 3611739942397549984 [TRT] Tactic: 3611739942397549984 Time: 0.599089 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148n_nt_v1 Tactic: 3827454225649558724 [TRT] Tactic: 3827454225649558724 Time: 0.458125 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v1 Tactic: 4337000649858996379 [TRT] Tactic: 4337000649858996379 Time: 0.610885 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.597317 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.590235 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.582812 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 5921334924264294896 [TRT] Tactic: 5921334924264294896 Time: 0.343021 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.603828 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.654401 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148t_nt_v1 Tactic: 7852627285308570038 [TRT] Tactic: 7852627285308570038 Time: 0.445834 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v1 Tactic: -9137461792520977713 [TRT] Tactic: -9137461792520977713 Time: 0.601823 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148n_nt_v0 Tactic: -8776506421218919509 [TRT] Tactic: -8776506421218919509 Time: 0.434636 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.602552 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v0 Tactic: -8133971918129952780 [TRT] Tactic: -8133971918129952780 Time: 0.642447 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v1 Tactic: -6092040395344634144 [TRT] Tactic: -6092040395344634144 Time: 0.777318 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.763906 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.633386 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148t_nt_v0 Tactic: -2318106587342035239 [TRT] Tactic: -2318106587342035239 Time: 0.428672 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_mobile_relu_tile148t_nt_v0 Tactic: -1343271414618805657 [TRT] Tactic: -1343271414618805657 Time: 0.314063 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.645546 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.614869 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.574766 [TRT] Fastest Tactic: -1343271414618805657 Time: 0.314063 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -1343271414618805657 [TRT] *************** Autotuning format combination: Float(59584,1,4256,304) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: inception_4a/3x3 + inception_4a/relu_3x3 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4a/3x3 + inception_4a/relu_3x3 (CaskConvolution) [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.657292 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.574193 [TRT] Fastest Tactic: -7394439838318485025 Time: 0.574193 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(59584,196,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/3x3 + inception_4a/relu_3x3 (CudnnConvolution) [TRT] Tactic: 0 Time: 1.17044 [TRT] Tactic: 1 Time: 1.17672 [TRT] Tactic: 2 Time: 1.01852 [TRT] Tactic: 4 skipped. Scratch requested: 46946304, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 88223744, available: 33554432 [TRT] Tactic: 6 Time: 1.66151 [TRT] Fastest Tactic: 2 Time: 1.01852 [TRT] --------------- Timing Runner: inception_4a/3x3 + inception_4a/relu_3x3 (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 2 [TRT] *************** Autotuning format combination: Half(29792,196:2,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/3x3 + inception_4a/relu_3x3 (FusedConvActConvolution) [TRT] Tactic: 524287 Time: 0.292943 [TRT] Tactic: 720895 Time: 0.41336 [TRT] Tactic: 983039 Time: 0.329454 [TRT] Tactic: 1048575 Time: 0.323802 [TRT] Tactic: 1703935 Time: 0.328256 [TRT] Tactic: 1769471 Time: 2.06948 [TRT] Tactic: 1966079 Time: 0.380338 [TRT] Tactic: 2031615 Time: 0.32487 [TRT] Tactic: 2228223 Time: 0.416042 [TRT] Tactic: 2424831 Time: 0.603516 [TRT] Tactic: 2621439 Time: 0.388411 [TRT] Tactic: 2752511 Time: 0.342943 [TRT] Tactic: 2818047 Time: 0.431432 [TRT] Tactic: 2883583 Time: 0.393907 [TRT] Tactic: 3014655 Time: 0.313307 [TRT] Tactic: 3145727 Time: 0.337994 [TRT] Tactic: 3473407 Time: 0.391954 [TRT] Tactic: 3604479 Time: 0.306041 [TRT] Tactic: 3735551 Time: 0.331329 [TRT] Tactic: 4390911 Time: 0.328593 [TRT] Tactic: 5046271 Time: 0.300677 [TRT] Tactic: 5963775 Time: 0.32362 [TRT] Tactic: 6160383 Time: 0.314454 [TRT] Tactic: 6488063 Time: 0.283152 [TRT] Tactic: 6881279 Time: 0.300808 [TRT] Tactic: 7274495 Time: 0.430209 [TRT] Tactic: 7864319 Time: 0.412214 [TRT] Tactic: 7995391 Time: 0.402604 [TRT] Tactic: 8585215 Time: 0.295443 [TRT] Tactic: 8847359 Time: 0.309922 [TRT] Tactic: 8978431 Time: 0.338516 [TRT] Tactic: 9043967 Time: 0.277291 [TRT] Tactic: 9175039 Time: 0.308932 [TRT] Tactic: 9502719 Time: 0.322865 [TRT] Tactic: 9830399 Time: 0.293593 [TRT] Tactic: 9961471 Time: 0.34138 [TRT] Tactic: 10027007 Time: 0.278099 [TRT] Tactic: 10092543 Time: 0.328229 [TRT] Tactic: 10289151 Time: 0.383125 [TRT] Tactic: 10485759 Time: 0.276744 [TRT] Tactic: 10682367 Time: 0.380417 [TRT] Tactic: 10813439 Time: 0.411745 [TRT] Fastest Tactic: 10485759 Time: 0.276744 [TRT] --------------- Timing Runner: inception_4a/3x3 + inception_4a/relu_3x3 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4a/3x3 + inception_4a/relu_3x3 (CaskConvolution) [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.373932 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_large_nn_v1 Tactic: 3650389455493082349 [TRT] Tactic: 3650389455493082349 Time: 0.390859 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_winograd_fp16x2_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 4772821744921268633 [TRT] Tactic: 4772821744921268633 Time: 0.187552 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.32401 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.319401 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_large_nn_v1 Tactic: -6490690591794140522 [TRT] Tactic: -6490690591794140522 Time: 0.32401 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_large_nn_v1 Tactic: -4686027666808657977 [TRT] Tactic: -4686027666808657977 Time: 0.308958 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.295364 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.307343 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.302162 [TRT] Fastest Tactic: 4772821744921268633 Time: 0.187552 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 4772821744921268633 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0732815 [TRT] Tactic: 0 Time: 0.066354 [TRT] Fastest Tactic: 0 Time: 0.066354 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.32854 [TRT] Tactic: 0 Time: 0.0609635 [TRT] Fastest Tactic: 0 Time: 0.0609635 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.070547 [TRT] Tactic: 0 Time: 0.036849 [TRT] Fastest Tactic: 0 Time: 0.036849 [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0866405 [TRT] Tactic: 0 Time: 0.0614585 [TRT] Fastest Tactic: 0 Time: 0.0614585 [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0541925 [TRT] Tactic: 0 Time: 0.060729 [TRT] Fastest Tactic: 1002 Time: 0.0541925 [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.061797 [TRT] Tactic: 0 Time: 0.0717445 [TRT] Fastest Tactic: 1002 Time: 0.061797 [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.35964 [TRT] Tactic: 0 Time: 0.0612495 [TRT] Fastest Tactic: 0 Time: 0.0612495 [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.050651 [TRT] Tactic: 0 Time: 0.06651 [TRT] Fastest Tactic: 1002 Time: 0.050651 [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0516665 [TRT] Tactic: 0 Time: 0.0379425 [TRT] Fastest Tactic: 0 Time: 0.0379425 [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0679165 [TRT] Tactic: 0 Time: 0.0706255 [TRT] Fastest Tactic: 1002 Time: 0.0679165 [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0507035 [TRT] Tactic: 0 Time: 0.0766405 [TRT] Fastest Tactic: 1002 Time: 0.0507035 [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.091354 [TRT] Tactic: 0 Time: 0.0636975 [TRT] Fastest Tactic: 0 Time: 0.0636975 [TRT] *************** Autotuning Reformat:Float(59584,196,14,1) -> Float(59584,1,4256,304) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0241145 [TRT] Tactic: 0 Time: 0.0079945 [TRT] Fastest Tactic: 0 Time: 0.0079945 [TRT] *************** Autotuning Reformat:Float(59584,196,14,1) -> Half(59584,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.107579 [TRT] Tactic: 0 Time: 0.007943 [TRT] Fastest Tactic: 0 Time: 0.007943 [TRT] *************** Autotuning Reformat:Float(59584,196,14,1) -> Half(29792,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.012578 [TRT] Tactic: 0 Time: 0.0091925 [TRT] Fastest Tactic: 0 Time: 0.0091925 [TRT] *************** Autotuning Reformat:Float(59584,1,4256,304) -> Float(59584,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0241665 [TRT] Tactic: 0 Time: 0.0079945 [TRT] Fastest Tactic: 0 Time: 0.0079945 [TRT] *************** Autotuning Reformat:Float(59584,1,4256,304) -> Half(59584,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.012682 [TRT] Tactic: 0 Time: 0.0079165 [TRT] Fastest Tactic: 0 Time: 0.0079165 [TRT] *************** Autotuning Reformat:Float(59584,1,4256,304) -> Half(29792,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.012656 [TRT] Tactic: 0 Time: 0.0091925 [TRT] Fastest Tactic: 0 Time: 0.0091925 [TRT] *************** Autotuning Reformat:Half(59584,196,14,1) -> Float(59584,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.109922 [TRT] Tactic: 0 Time: 0.0055205 [TRT] Fastest Tactic: 0 Time: 0.0055205 [TRT] *************** Autotuning Reformat:Half(59584,196,14,1) -> Float(59584,1,4256,304) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0115885 [TRT] Tactic: 0 Time: 0.0079165 [TRT] Fastest Tactic: 0 Time: 0.0079165 [TRT] *************** Autotuning Reformat:Half(59584,196,14,1) -> Half(29792,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.012474 [TRT] Tactic: 0 Time: 0.005495 [TRT] Fastest Tactic: 0 Time: 0.005495 [TRT] *************** Autotuning Reformat:Half(29792,196:2,14,1) -> Float(59584,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0126825 [TRT] Tactic: 0 Time: 0.005495 [TRT] Fastest Tactic: 0 Time: 0.005495 [TRT] *************** Autotuning Reformat:Half(29792,196:2,14,1) -> Float(59584,1,4256,304) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0125255 [TRT] Tactic: 0 Time: 0.010209 [TRT] Fastest Tactic: 0 Time: 0.010209 [TRT] *************** Autotuning Reformat:Half(29792,196:2,14,1) -> Half(59584,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0220315 [TRT] Tactic: 0 Time: 0.0079685 [TRT] Fastest Tactic: 0 Time: 0.0079685 [TRT] *************** Autotuning format combination: Float(59584,196,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/5x5 + inception_4a/relu_5x5 (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4a/5x5 + inception_4a/relu_5x5 (FusedConvActConvolution) [TRT] Tactic: 393215 Time: 0.0966925 [TRT] Tactic: 917503 Time: 0.087734 [TRT] Tactic: 1114111 Time: 0.0903125 [TRT] Tactic: 1245183 Time: 0.0924475 [TRT] Tactic: 1572863 Time: 0.0938545 [TRT] Tactic: 2490367 Time: 0.110651 [TRT] Tactic: 2555903 Time: 0.105312 [TRT] Tactic: 2949119 Time: 0.080547 [TRT] Tactic: 3211263 Time: 0.15526 [TRT] Tactic: 3801087 Time: 0.09151 [TRT] Tactic: 3866623 Time: 0.0949995 [TRT] Tactic: 4128767 Time: 0.078229 [TRT] Tactic: 4456447 Time: 0.085 [TRT] Tactic: 4718591 Time: 0.0814845 [TRT] Tactic: 4784127 Time: 0.203698 [TRT] Tactic: 4849663 Time: 0.082682 [TRT] Tactic: 5111807 Time: 0.082109 [TRT] Tactic: 5308415 Time: 0.0963805 [TRT] Tactic: 5505023 Time: 0.137682 [TRT] Tactic: 6094847 Time: 0.0950255 [TRT] Tactic: 6356991 Time: 0.104245 [TRT] Tactic: 6553599 Time: 0.095989 [TRT] Tactic: 6619135 Time: 0.103854 [TRT] Tactic: 6684671 Time: 0.146901 [TRT] Tactic: 7471103 Time: 0.094349 [TRT] Tactic: 7667711 Time: 0.081016 [TRT] Tactic: 7929855 Time: 0.0883855 [TRT] Tactic: 8060927 Time: 0.101172 [TRT] Tactic: 8126463 Time: 0.0873695 [TRT] Tactic: 8388607 Time: 0.0960155 [TRT] Tactic: 8519679 Time: 0.109088 [TRT] Tactic: 8781823 Time: 0.100599 [TRT] Tactic: 8912895 Time: 0.087005 [TRT] Tactic: 9240575 Time: 0.078776 [TRT] Tactic: 9306111 Time: 0.133099 [TRT] Tactic: 9371647 Time: 0.0845315 [TRT] Tactic: 9437183 Time: 0.0804945 [TRT] Tactic: 9633791 Time: 0.0848435 [TRT] Tactic: 9699327 Time: 0.098438 [TRT] Tactic: 9764863 Time: 0.103281 [TRT] Tactic: 10158079 Time: 0.100364 [TRT] Tactic: 10420223 Time: 0.0915885 [TRT] Tactic: 10616831 Time: 0.091875 [TRT] Tactic: 10878975 Time: 0.107474 [TRT] Fastest Tactic: 4128767 Time: 0.078229 [TRT] --------------- Timing Runner: inception_4a/5x5 + inception_4a/relu_5x5 (CudnnConvolution) [TRT] Tactic: 0 Time: 0.197136 [TRT] Tactic: 1 Time: 0.197187 [TRT] Tactic: 2 Time: 0.322995 [TRT] Tactic: 4 Time: 1.63115 [TRT] Tactic: 5 Time: 1.66477 [TRT] Fastest Tactic: 0 Time: 0.197136 [TRT] --------------- Timing Runner: inception_4a/5x5 + inception_4a/relu_5x5 (CaskConvolution) [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.135521 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v0 Tactic: 1754984623894446479 [TRT] Tactic: 1754984623894446479 Time: 0.12763 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v0 Tactic: 3611739942397549984 [TRT] Tactic: 3611739942397549984 Time: 0.155443 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v1 Tactic: 4337000649858996379 [TRT] Tactic: 4337000649858996379 Time: 0.108047 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.153958 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.0904165 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.146041 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.10651 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.098307 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v1 Tactic: -9137461792520977713 [TRT] Tactic: -9137461792520977713 Time: 0.156562 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.157501 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v0 Tactic: -8133971918129952780 [TRT] Tactic: -8133971918129952780 Time: 0.107917 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v1 Tactic: -6092040395344634144 [TRT] Tactic: -6092040395344634144 Time: 0.138256 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.126145 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.0995055 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.106432 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.0928385 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.145781 [TRT] Fastest Tactic: 5137655947464784826 Time: 0.0904165 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 4128767 [TRT] *************** Autotuning format combination: Float(59584,1,4256,304) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: inception_4a/5x5 + inception_4a/relu_5x5 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4a/5x5 + inception_4a/relu_5x5 (CaskConvolution) [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.184036 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.078724 [TRT] Fastest Tactic: -7394439838318485025 Time: 0.078724 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(59584,196,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/5x5 + inception_4a/relu_5x5 (CudnnConvolution) [TRT] Tactic: 0 Time: 0.193021 [TRT] Tactic: 1 Time: 0.384115 [TRT] Tactic: 2 Time: 0.313932 [TRT] Tactic: 4 Time: 1.61177 [TRT] Tactic: 5 Time: 1.66492 [TRT] Fastest Tactic: 0 Time: 0.193021 [TRT] --------------- Timing Runner: inception_4a/5x5 + inception_4a/relu_5x5 (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 0 [TRT] *************** Autotuning format combination: Half(29792,196:2,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/5x5 + inception_4a/relu_5x5 (FusedConvActConvolution) [TRT] Tactic: 393215 Time: 0.056745 [TRT] Tactic: 917503 Time: 0.055 [TRT] Tactic: 1114111 Time: 0.0504945 [TRT] Tactic: 1245183 Time: 0.0497395 [TRT] Tactic: 1572863 Time: 0.059297 [TRT] Tactic: 2490367 Time: 0.0786975 [TRT] Tactic: 2555903 Time: 0.0775 [TRT] Tactic: 2949119 Time: 0.053672 [TRT] Tactic: 3211263 Time: 0.0972135 [TRT] Tactic: 3801087 Time: 0.054688 [TRT] Tactic: 3866623 Time: 0.0567185 [TRT] Tactic: 4128767 Time: 0.0490885 [TRT] Tactic: 4456447 Time: 0.054193 [TRT] Tactic: 4718591 Time: 0.047109 [TRT] Tactic: 4784127 Time: 0.106354 [TRT] Tactic: 4849663 Time: 0.0478125 [TRT] Tactic: 5111807 Time: 0.047865 [TRT] Tactic: 5308415 Time: 0.057474 [TRT] Tactic: 5505023 Time: 0.0869795 [TRT] Tactic: 6094847 Time: 0.0567185 [TRT] Tactic: 6356991 Time: 0.0751305 [TRT] Tactic: 6553599 Time: 0.057083 [TRT] Tactic: 6619135 Time: 0.055443 [TRT] Tactic: 6684671 Time: 0.093724 [TRT] Tactic: 7471103 Time: 0.056771 [TRT] Tactic: 7667711 Time: 0.0477865 [TRT] Tactic: 7929855 Time: 0.0480465 [TRT] Tactic: 8060927 Time: 0.0636975 [TRT] Tactic: 8126463 Time: 0.0458335 [TRT] Tactic: 8388607 Time: 0.0599215 [TRT] Tactic: 8519679 Time: 0.0641405 [TRT] Tactic: 8781823 Time: 0.0534115 [TRT] Tactic: 8912895 Time: 0.054687 [TRT] Tactic: 9240575 Time: 0.044766 [TRT] Tactic: 9306111 Time: 0.068255 [TRT] Tactic: 9371647 Time: 0.0470055 [TRT] Tactic: 9437183 Time: 0.053099 [TRT] Tactic: 9633791 Time: 0.048203 [TRT] Tactic: 9699327 Time: 0.0590625 [TRT] Tactic: 9764863 Time: 0.0646615 [TRT] Tactic: 10158079 Time: 0.060599 [TRT] Tactic: 10420223 Time: 0.061484 [TRT] Tactic: 10616831 Time: 0.054896 [TRT] Tactic: 10878975 Time: 0.067136 [TRT] Fastest Tactic: 9240575 Time: 0.044766 [TRT] --------------- Timing Runner: inception_4a/5x5 + inception_4a/relu_5x5 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4a/5x5 + inception_4a/relu_5x5 (CaskConvolution) [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.075208 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_large_nn_v1 Tactic: 3650389455493082349 [TRT] Tactic: 3650389455493082349 Time: 0.0771615 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.0566145 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.061302 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_large_nn_v1 Tactic: -6490690591794140522 [TRT] Tactic: -6490690591794140522 Time: 0.061667 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_large_nn_v1 Tactic: -4686027666808657977 [TRT] Tactic: -4686027666808657977 Time: 0.0859375 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.0785675 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.0835415 [TRT] inception_4a/5x5 + inception_4a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.048438 [TRT] Fastest Tactic: -2409163523992614473 Time: 0.048438 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 9240575 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0242705 [TRT] Tactic: 0 Time: 0.017266 [TRT] Fastest Tactic: 0 Time: 0.017266 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.311459 [TRT] Tactic: 0 Time: 0.01711 [TRT] Fastest Tactic: 0 Time: 0.01711 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0217705 [TRT] Tactic: 0 Time: 0.0105465 [TRT] Fastest Tactic: 0 Time: 0.0105465 [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.029271 [TRT] Tactic: 0 Time: 0.017422 [TRT] Fastest Tactic: 0 Time: 0.017422 [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.018229 [TRT] Tactic: 0 Time: 0.017656 [TRT] Fastest Tactic: 0 Time: 0.017656 [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0201305 [TRT] Tactic: 0 Time: 0.019479 [TRT] Fastest Tactic: 0 Time: 0.019479 [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.319245 [TRT] Tactic: 0 Time: 0.0171355 [TRT] Fastest Tactic: 0 Time: 0.0171355 [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.017344 [TRT] Tactic: 0 Time: 0.0184895 [TRT] Fastest Tactic: 1002 Time: 0.017344 [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.017344 [TRT] Tactic: 0 Time: 0.0104685 [TRT] Fastest Tactic: 0 Time: 0.0104685 [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.021901 [TRT] Tactic: 0 Time: 0.0195575 [TRT] Fastest Tactic: 0 Time: 0.0195575 [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.019609 [TRT] Tactic: 0 Time: 0.019661 [TRT] Fastest Tactic: 1002 Time: 0.019609 [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0300515 [TRT] Tactic: 0 Time: 0.0171095 [TRT] Fastest Tactic: 0 Time: 0.0171095 [TRT] *************** Autotuning Reformat:Float(94080,196,14,1) -> Half(94080,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(94080,196,14,1) -> Half(47040,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Float(94080,1,6720,480) -> Float(94080,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(94080,1,6720,480) -> Half(94080,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(94080,1,6720,480) -> Half(47040,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(94080,196,14,1) -> Float(94080,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(94080,196,14,1) -> Half(47040,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(47040,196:2,14,1) -> Float(94080,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(47040,196:2,14,1) -> Half(94080,196,14,1) *************** [TRT] *************** Autotuning format combination: Float(94080,196,14,1) -> Float(94080,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/pool (TiledPooling) [TRT] Tactic: 2752769 Time: 0.200859 [TRT] Tactic: 2818305 Time: 0.190807 [TRT] Tactic: 2883841 Time: 0.142265 [TRT] Tactic: 2949377 Time: 0.772942 [TRT] Tactic: 3014913 Time: 0.695781 [TRT] Tactic: 3080449 Time: 0.383047 [TRT] Tactic: 3145985 Time: 0.309401 [TRT] Tactic: 3211521 Time: 0.129166 [TRT] Tactic: 3277057 Time: 0.124818 [TRT] Tactic: 3342593 Time: 0.097188 [TRT] Tactic: 3408129 Time: 0.447682 [TRT] Tactic: 3473665 Time: 0.404167 [TRT] Tactic: 3539201 Time: 0.235052 [TRT] Tactic: 3604737 Time: 0.189688 [TRT] Tactic: 3670273 Time: 0.113594 [TRT] Tactic: 3735809 Time: 0.111301 [TRT] Tactic: 3801345 Time: 0.092135 [TRT] Tactic: 3866881 Time: 0.384402 [TRT] Tactic: 3932417 Time: 0.339688 [TRT] Tactic: 3997953 Time: 0.208828 [TRT] Tactic: 4063489 Time: 0.169531 [TRT] Tactic: 4129025 Time: 0.103594 [TRT] Tactic: 4194561 Time: 0.101615 [TRT] Tactic: 4260097 Time: 0.086641 [TRT] Tactic: 4325633 Time: 0.329089 [TRT] Tactic: 4391169 Time: 0.290704 [TRT] Tactic: 4456705 Time: 0.183594 [TRT] Tactic: 4522241 Time: 0.150078 [TRT] Tactic: 4587777 Time: 0.102709 [TRT] Tactic: 4653313 Time: 0.102578 [TRT] Tactic: 4718849 Time: 0.084323 [TRT] Tactic: 4784385 Time: 0.304662 [TRT] Tactic: 4849921 Time: 0.274219 [TRT] Tactic: 4915457 Time: 0.174271 [TRT] Tactic: 4980993 Time: 0.139245 [TRT] Tactic: 5046529 Time: 0.104271 [TRT] Tactic: 5112065 Time: 0.105052 [TRT] Tactic: 5177601 Time: 0.0826565 [TRT] Tactic: 5243137 Time: 0.283125 [TRT] Tactic: 5308673 Time: 0.254531 [TRT] Tactic: 5374209 Time: 0.167266 [TRT] Tactic: 5439745 Time: 0.133047 [TRT] Tactic: 6553857 Time: 0.166224 [TRT] Tactic: 6750465 Time: 0.112135 [TRT] Fastest Tactic: 5177601 Time: 0.0826565 [TRT] --------------- Timing Runner: inception_4a/pool (CudnnPooling) [TRT] Tactic: -1 Time: 0.328438 [TRT] Fastest Tactic: -1 Time: 0.328438 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: TiledPooling Tactic: 5177601 [TRT] *************** Autotuning format combination: Half(94080,196,14,1) -> Half(94080,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/pool (TiledPooling) [TRT] TiledPooling has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4a/pool (CudnnPooling) [TRT] Tactic: -1 Time: 0.342526 [TRT] Fastest Tactic: -1 Time: 0.342526 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnPooling Tactic: -1 [TRT] *************** Autotuning format combination: Half(47040,196:2,14,1) -> Half(47040,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/pool (TiledPooling) [TRT] Tactic: 2752769 Time: 0.123463 [TRT] Tactic: 2818305 Time: 0.12263 [TRT] Tactic: 2883841 Time: 0.0929165 [TRT] Tactic: 2949377 Time: 0.459115 [TRT] Tactic: 3014913 Time: 0.426588 [TRT] Tactic: 3080449 Time: 0.241589 [TRT] Tactic: 3145985 Time: 0.190729 [TRT] Tactic: 3211521 Time: 0.088984 [TRT] Tactic: 3277057 Time: 0.0869535 [TRT] Tactic: 3342593 Time: 0.0688285 [TRT] Tactic: 3408129 Time: 0.308411 [TRT] Tactic: 3473665 Time: 0.279454 [TRT] Tactic: 3539201 Time: 0.16375 [TRT] Tactic: 3604737 Time: 0.136719 [TRT] Tactic: 3670273 Time: 0.076875 [TRT] Tactic: 3735809 Time: 0.074427 [TRT] Tactic: 3801345 Time: 0.060261 [TRT] Tactic: 3866881 Time: 0.262995 [TRT] Tactic: 3932417 Time: 0.23724 [TRT] Tactic: 3997953 Time: 0.141693 [TRT] Tactic: 4063489 Time: 0.118802 [TRT] Tactic: 4129025 Time: 0.0694275 [TRT] Tactic: 4194561 Time: 0.070208 [TRT] Tactic: 4260097 Time: 0.0572395 [TRT] Tactic: 4325633 Time: 0.233386 [TRT] Tactic: 4391169 Time: 0.217995 [TRT] Tactic: 4456705 Time: 0.131041 [TRT] Tactic: 4522241 Time: 0.109115 [TRT] Tactic: 4587777 Time: 0.0677605 [TRT] Tactic: 4653313 Time: 0.066953 [TRT] Tactic: 4718849 Time: 0.0548955 [TRT] Tactic: 4784385 Time: 0.222214 [TRT] Tactic: 4849921 Time: 0.205833 [TRT] Tactic: 4915457 Time: 0.124062 [TRT] Tactic: 4980993 Time: 0.105156 [TRT] Tactic: 5046529 Time: 0.0666925 [TRT] Tactic: 5112065 Time: 0.0661715 [TRT] Tactic: 5177601 Time: 0.054323 [TRT] Tactic: 5243137 Time: 0.214922 [TRT] Tactic: 5308673 Time: 0.197735 [TRT] Tactic: 5374209 Time: 0.119584 [TRT] Tactic: 5439745 Time: 0.102318 [TRT] Tactic: 6553857 Time: 0.104531 [TRT] Tactic: 6750465 Time: 0.0763545 [TRT] Fastest Tactic: 5177601 Time: 0.054323 [TRT] --------------- Timing Runner: inception_4a/pool (CudaPooling) [TRT] Tactic: -3 Time: 0.185677 [TRT] Fastest Tactic: -3 Time: 0.185677 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: TiledPooling Tactic: 5177601 [TRT] *************** Autotuning Reformat:Float(94080,196,14,1) -> Float(94080,1,6720,480) *************** [TRT] *************** Autotuning Reformat:Float(94080,196,14,1) -> Half(94080,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(94080,196,14,1) -> Half(47040,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Float(94080,1,6720,480) -> Float(94080,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(94080,1,6720,480) -> Half(94080,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(94080,1,6720,480) -> Half(47040,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(94080,196,14,1) -> Float(94080,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(94080,196,14,1) -> Float(94080,1,6720,480) *************** [TRT] *************** Autotuning Reformat:Half(94080,196,14,1) -> Half(47040,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(47040,196:2,14,1) -> Float(94080,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(47040,196:2,14,1) -> Float(94080,1,6720,480) *************** [TRT] *************** Autotuning Reformat:Half(47040,196:2,14,1) -> Half(94080,196,14,1) *************** [TRT] *************** Autotuning format combination: Float(94080,196,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/pool_proj + inception_4a/relu_pool_proj (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4a/pool_proj + inception_4a/relu_pool_proj (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.213177 [TRT] Tactic: 655359 Time: 0.240885 [TRT] Tactic: 786431 Time: 0.205391 [TRT] Tactic: 851967 Time: 0.339557 [TRT] Tactic: 1179647 Time: 0.187969 [TRT] Tactic: 1310719 Time: 0.424844 [TRT] Tactic: 1376255 Time: 0.225677 [TRT] Tactic: 1441791 Time: 0.238359 [TRT] Tactic: 1507327 Time: 0.443698 [TRT] Tactic: 1638399 Time: 0.281016 [TRT] Tactic: 1835007 Time: 0.216771 [TRT] Tactic: 1900543 Time: 0.428281 [TRT] Tactic: 2097151 Time: 0.22776 [TRT] Tactic: 2162687 Time: 0.234974 [TRT] Tactic: 2293759 Time: 0.228568 [TRT] Tactic: 2359295 Time: 0.246588 [TRT] Tactic: 2686975 Time: 0.228828 [TRT] Tactic: 3080191 Time: 0.271329 [TRT] Tactic: 3342335 Time: 0.352214 [TRT] Tactic: 3407871 Time: 0.173281 [TRT] Tactic: 3538943 Time: 0.167396 [TRT] Tactic: 3670015 Time: 0.274844 [TRT] Tactic: 3932159 Time: 0.384401 [TRT] Tactic: 3997695 Time: 0.208932 [TRT] Tactic: 4063231 Time: 0.296589 [TRT] Tactic: 4194303 Time: 0.206849 [TRT] Tactic: 4259839 Time: 0.237031 [TRT] Tactic: 4325375 Time: 0.264583 [TRT] Tactic: 4521983 Time: 0.258489 [TRT] Tactic: 4587519 Time: 0.232187 [TRT] Tactic: 4653055 Time: 0.207839 [TRT] Tactic: 4915199 Time: 0.217031 [TRT] Tactic: 4980735 Time: 0.312708 [TRT] Tactic: 5177343 Time: 0.222995 [TRT] Tactic: 5242879 Time: 0.155026 [TRT] Tactic: 5373951 Time: 0.212474 [TRT] Tactic: 5439487 Time: 0.204167 [TRT] Tactic: 5570559 Time: 0.225989 [TRT] Tactic: 5636095 Time: 0.295729 [TRT] Tactic: 5701631 Time: 0.226354 [TRT] Tactic: 5767167 Time: 0.28388 [TRT] Tactic: 5832703 Time: 0.17052 [TRT] Tactic: 5898239 Time: 0.154297 [TRT] Tactic: 6029311 Time: 0.215912 [TRT] Tactic: 6225919 Time: 0.150547 [TRT] Tactic: 6291455 Time: 0.188594 [TRT] Tactic: 6422527 Time: 0.288072 [TRT] Tactic: 6750207 Time: 0.205026 [TRT] Tactic: 6815743 Time: 0.157109 [TRT] Tactic: 6946815 Time: 0.290105 [TRT] Tactic: 7012351 Time: 0.23 [TRT] Tactic: 7077887 Time: 0.165338 [TRT] Tactic: 7143423 Time: 0.29974 [TRT] Tactic: 7208959 Time: 0.22987 [TRT] Tactic: 7340031 Time: 0.164323 [TRT] Tactic: 7405567 Time: 0.188803 [TRT] Tactic: 7536639 Time: 0.179896 [TRT] Tactic: 7602175 Time: 0.275443 [TRT] Tactic: 7733247 Time: 0.158906 [TRT] Tactic: 7798783 Time: 0.205573 [TRT] Tactic: 8191999 Time: 0.306901 [TRT] Tactic: 8257535 Time: 0.22177 [TRT] Tactic: 8323071 Time: 0.196536 [TRT] Tactic: 8650751 Time: 0.288541 [TRT] Tactic: 8716287 Time: 0.157526 [TRT] Tactic: 9109503 Time: 0.24586 [TRT] Tactic: 9568255 Time: 0.217526 [TRT] Tactic: 9895935 Time: 0.205885 [TRT] Tactic: 10223615 Time: 0.230494 [TRT] Tactic: 10354687 Time: 0.250365 [TRT] Tactic: 10551295 Time: 0.184218 [TRT] Tactic: 10747903 Time: 0.15823 [TRT] Tactic: 10944511 Time: 0.313021 [TRT] Fastest Tactic: 6225919 Time: 0.150547 [TRT] --------------- Timing Runner: inception_4a/pool_proj + inception_4a/relu_pool_proj (CudnnConvolution) [TRT] Tactic: 0 Time: 0.243125 [TRT] Tactic: 1 Time: 0.206744 [TRT] Tactic: 2 Time: 0.520182 [TRT] Tactic: 4 skipped. Scratch requested: 72007680, available: 33554432 [TRT] Tactic: 5 Time: 2.08471 [TRT] Fastest Tactic: 1 Time: 0.206744 [TRT] --------------- Timing Runner: inception_4a/pool_proj + inception_4a/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4a/pool_proj + inception_4a/relu_pool_proj (CaskConvolution) [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.147005 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v0 Tactic: 1698681053543049347 [TRT] Tactic: 1698681053543049347 Time: 0.131745 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.188646 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.115938 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.186562 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v1 Tactic: 5326823351883942011 [TRT] Tactic: 5326823351883942011 Time: 0.181901 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v0 Tactic: 5500448035057547314 [TRT] Tactic: 5500448035057547314 Time: 0.145157 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.122005 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.137552 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.190078 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v1 Tactic: -6576203419454146580 [TRT] Tactic: -6576203419454146580 Time: 0.130652 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.143646 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.140235 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.15224 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.148333 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.186641 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v0 Tactic: -377491875521947884 [TRT] Tactic: -377491875521947884 Time: 0.186016 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v1 Tactic: -37215280111360163 [TRT] Tactic: -37215280111360163 Time: 0.113568 [TRT] Fastest Tactic: -37215280111360163 Time: 0.113568 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -37215280111360163 [TRT] *************** Autotuning format combination: Float(94080,1,6720,480) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: inception_4a/pool_proj + inception_4a/relu_pool_proj (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4a/pool_proj + inception_4a/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4a/pool_proj + inception_4a/relu_pool_proj (CaskConvolution) [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 3886731678879822788 [TRT] Tactic: 3886731678879822788 Time: 0.100547 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 6629944304117643200 [TRT] Tactic: 6629944304117643200 Time: 0.157553 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.158125 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.101406 [TRT] Fastest Tactic: 3886731678879822788 Time: 0.100547 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 3886731678879822788 [TRT] *************** Autotuning format combination: Half(94080,196,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/pool_proj + inception_4a/relu_pool_proj (CudnnConvolution) [TRT] Tactic: 0 Time: 0.247995 [TRT] Tactic: 1 Time: 0.237864 [TRT] Tactic: 2 Time: 0.515442 [TRT] Tactic: 4 skipped. Scratch requested: 72007680, available: 33554432 [TRT] Tactic: 5 Time: 1.97162 [TRT] Fastest Tactic: 1 Time: 0.237864 [TRT] --------------- Timing Runner: inception_4a/pool_proj + inception_4a/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4a/pool_proj + inception_4a/relu_pool_proj (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 1 [TRT] *************** Autotuning format combination: Half(47040,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/pool_proj + inception_4a/relu_pool_proj (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] *************** Autotuning format combination: Half(47040,196:2,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/pool_proj + inception_4a/relu_pool_proj (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.11375 [TRT] Tactic: 655359 Time: 0.188775 [TRT] Tactic: 786431 Time: 0.142526 [TRT] Tactic: 851967 Time: 0.175807 [TRT] Tactic: 1179647 Time: 0.090495 [TRT] Tactic: 1310719 Time: 0.245885 [TRT] Tactic: 1376255 Time: 0.11664 [TRT] Tactic: 1441791 Time: 0.122917 [TRT] Tactic: 1507327 Time: 0.21948 [TRT] Tactic: 1638399 Time: 0.148594 [TRT] Tactic: 1835007 Time: 0.137812 [TRT] Tactic: 1900543 Time: 0.20711 [TRT] Tactic: 2162687 Time: 0.116849 [TRT] Tactic: 2293759 Time: 0.117916 [TRT] Tactic: 2359295 Time: 0.119557 [TRT] Tactic: 2686975 Time: 0.207318 [TRT] Tactic: 3080191 Time: 0.151693 [TRT] Tactic: 3342335 Time: 0.174766 [TRT] Tactic: 3407871 Time: 0.101016 [TRT] Tactic: 3538943 Time: 0.090573 [TRT] Tactic: 3670015 Time: 0.20849 [TRT] Tactic: 3932159 Time: 0.165156 [TRT] Tactic: 3997695 Time: 0.130469 [TRT] Tactic: 4063231 Time: 0.156927 [TRT] Tactic: 4194303 Time: 0.108854 [TRT] Tactic: 4325375 Time: 0.132084 [TRT] Tactic: 4521983 Time: 0.135703 [TRT] Tactic: 4587519 Time: 0.133308 [TRT] Tactic: 4653055 Time: 0.110105 [TRT] Tactic: 4915199 Time: 0.110963 [TRT] Tactic: 4980735 Time: 0.14164 [TRT] Tactic: 5177343 Time: 0.100728 [TRT] Tactic: 5242879 Time: 0.080313 [TRT] Tactic: 5373951 Time: 0.093359 [TRT] Tactic: 5439487 Time: 0.108125 [TRT] Tactic: 5570559 Time: 0.168047 [TRT] Tactic: 5636095 Time: 0.156927 [TRT] Tactic: 5701631 Time: 0.111119 [TRT] Tactic: 5767167 Time: 0.141979 [TRT] Tactic: 5832703 Time: 0.095052 [TRT] Tactic: 5898239 Time: 0.0910675 [TRT] Tactic: 6029311 Time: 0.107916 [TRT] Tactic: 6225919 Time: 0.0777345 [TRT] Tactic: 6291455 Time: 0.090469 [TRT] Tactic: 6422527 Time: 0.143203 [TRT] Tactic: 6750207 Time: 0.107682 [TRT] Tactic: 6815743 Time: 0.0829165 [TRT] Tactic: 6946815 Time: 0.130183 [TRT] Tactic: 7077887 Time: 0.0852345 [TRT] Tactic: 7143423 Time: 0.15388 [TRT] Tactic: 7208959 Time: 0.100235 [TRT] Tactic: 7340031 Time: 0.0941665 [TRT] Tactic: 7405567 Time: 0.0902085 [TRT] Tactic: 7536639 Time: 0.109687 [TRT] Tactic: 7602175 Time: 0.122057 [TRT] Tactic: 7733247 Time: 0.089453 [TRT] Tactic: 7798783 Time: 0.142578 [TRT] Tactic: 8191999 Time: 0.158776 [TRT] Tactic: 8257535 Time: 0.109141 [TRT] Tactic: 8323071 Time: 0.104687 [TRT] Tactic: 8650751 Time: 0.132188 [TRT] Tactic: 8716287 Time: 0.0803905 [TRT] Tactic: 9568255 Time: 0.111042 [TRT] Tactic: 9895935 Time: 0.108464 [TRT] Tactic: 10223615 Time: 0.204662 [TRT] Tactic: 10354687 Time: 0.141484 [TRT] Tactic: 10551295 Time: 0.098828 [TRT] Tactic: 10747903 Time: 0.0834895 [TRT] Tactic: 10944511 Time: 0.140938 [TRT] Fastest Tactic: 6225919 Time: 0.0777345 [TRT] --------------- Timing Runner: inception_4a/pool_proj + inception_4a/relu_pool_proj (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4a/pool_proj + inception_4a/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4a/pool_proj + inception_4a/relu_pool_proj (CaskConvolution) [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_interior_nn_v1 Tactic: 3066127711859985668 [TRT] Tactic: 3066127711859985668 Time: 0.0683855 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.07349 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.0691405 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.069115 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] Tactic: 8163473458334948789 Time: 0.0612765 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.098802 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.101562 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.0644795 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_interior_nn_v1 Tactic: -1716393687483585322 [TRT] Tactic: -1716393687483585322 Time: 0.0985425 [TRT] Fastest Tactic: 8163473458334948789 Time: 0.0612765 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 8163473458334948789 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0252865 [TRT] Tactic: 0 Time: 0.0240105 [TRT] Fastest Tactic: 0 Time: 0.0240105 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.449896 [TRT] Tactic: 0 Time: 0.0231255 [TRT] Fastest Tactic: 0 Time: 0.0231255 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.028099 [TRT] Tactic: 0 Time: 0.0157555 [TRT] Fastest Tactic: 0 Time: 0.0157555 [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.033359 [TRT] Tactic: 0 Time: 0.0233595 [TRT] Fastest Tactic: 0 Time: 0.0233595 [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0214845 [TRT] Tactic: 0 Time: 0.02375 [TRT] Fastest Tactic: 1002 Time: 0.0214845 [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.025859 [TRT] Tactic: 0 Time: 0.026537 [TRT] Fastest Tactic: 1002 Time: 0.025859 [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.460651 [TRT] Tactic: 0 Time: 0.0233075 [TRT] Fastest Tactic: 0 Time: 0.0233075 [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.018724 [TRT] Tactic: 0 Time: 0.026406 [TRT] Fastest Tactic: 1002 Time: 0.018724 [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.026068 [TRT] Tactic: 0 Time: 0.0152085 [TRT] Fastest Tactic: 0 Time: 0.0152085 [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.028021 [TRT] Tactic: 0 Time: 0.0260415 [TRT] Fastest Tactic: 0 Time: 0.0260415 [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.01862 [TRT] Tactic: 0 Time: 0.0283075 [TRT] Fastest Tactic: 1002 Time: 0.01862 [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.036641 [TRT] Tactic: 0 Time: 0.023516 [TRT] Fastest Tactic: 0 Time: 0.023516 [TRT] *************** Autotuning Reformat:Float(59584,196,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 3.0768 [TRT] Tactic: 0 Time: 0.0235675 [TRT] Fastest Tactic: 0 Time: 0.0235675 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(59584,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: inception_4a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.056823 [TRT] Tactic: 0 Time: 0.0660675 [TRT] Fastest Tactic: 1002 Time: 0.056823 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(59584,196,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 1.33841 [TRT] Tactic: 0 Time: 0.060156 [TRT] Fastest Tactic: 0 Time: 0.060156 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(59584,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0715105 [TRT] Tactic: 0 Time: 0.0721875 [TRT] Fastest Tactic: 1002 Time: 0.0715105 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(59584,1,4256,304) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.073646 [TRT] Tactic: 0 Time: 0.0618745 [TRT] Fastest Tactic: 0 Time: 0.0618745 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(59584,1,4256,304) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: inception_4a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.103437 [TRT] Tactic: 0 Time: 0.0660415 [TRT] Fastest Tactic: 0 Time: 0.0660415 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(59584,1,4256,304) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0515625 [TRT] Tactic: 0 Time: 0.062422 [TRT] Fastest Tactic: 1002 Time: 0.0515625 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(59584,1,4256,304) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0621875 [TRT] Tactic: 0 Time: 0.0727085 [TRT] Fastest Tactic: 1002 Time: 0.0621875 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(59584,196,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 1.36906 [TRT] Tactic: 0 Time: 0.061562 [TRT] Fastest Tactic: 0 Time: 0.061562 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(59584,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: inception_4a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.04737 [TRT] Tactic: 0 Time: 0.067344 [TRT] Fastest Tactic: 1002 Time: 0.04737 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(59584,196,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 1.36289 [TRT] Tactic: 0 Time: 0.008437 [TRT] Fastest Tactic: 0 Time: 0.008437 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(59584,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.051771 [TRT] Tactic: 0 Time: 0.036823 [TRT] Fastest Tactic: 0 Time: 0.036823 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(29792,196:2,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.067526 [TRT] Tactic: 0 Time: 0.0715885 [TRT] Fastest Tactic: 1002 Time: 0.067526 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(29792,196:2,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: inception_4a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0473695 [TRT] Tactic: 0 Time: 0.078073 [TRT] Fastest Tactic: 1002 Time: 0.0473695 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(29792,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0928385 [TRT] Tactic: 0 Time: 0.0647915 [TRT] Fastest Tactic: 0 Time: 0.0647915 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(29792,196:2,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.06888 [TRT] Tactic: 0 Time: 0.008412 [TRT] Fastest Tactic: 0 Time: 0.008412 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.146953 [TRT] Tactic: 0 Time: 0.167109 [TRT] Fastest Tactic: 1002 Time: 0.146953 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 3.19307 [TRT] Tactic: 0 Time: 0.128619 [TRT] Fastest Tactic: 0 Time: 0.128619 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.196537 [TRT] Tactic: 0 Time: 0.103255 [TRT] Fastest Tactic: 0 Time: 0.103255 [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.195677 [TRT] Tactic: 0 Time: 0.170312 [TRT] Fastest Tactic: 0 Time: 0.170312 [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.139037 [TRT] Tactic: 0 Time: 0.171589 [TRT] Fastest Tactic: 1002 Time: 0.139037 [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.169896 [TRT] Tactic: 0 Time: 0.203542 [TRT] Fastest Tactic: 1002 Time: 0.169896 [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 2.78203 [TRT] Tactic: 0 Time: 0.092448 [TRT] Fastest Tactic: 0 Time: 0.092448 [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.108124 [TRT] Tactic: 0 Time: 0.155782 [TRT] Fastest Tactic: 1002 Time: 0.108124 [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.121979 [TRT] Tactic: 0 Time: 0.0859115 [TRT] Fastest Tactic: 0 Time: 0.0859115 [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.15237 [TRT] Tactic: 0 Time: 0.0761465 [TRT] Fastest Tactic: 0 Time: 0.0761465 [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.10875 [TRT] Tactic: 0 Time: 0.180339 [TRT] Fastest Tactic: 1002 Time: 0.10875 [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.215625 [TRT] Tactic: 0 Time: 0.0735155 [TRT] Fastest Tactic: 0 Time: 0.0735155 [TRT] *************** Autotuning format combination: Float(100352,196,14,1) -> Float(58016,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.880625 [TRT] Tactic: 655359 Time: 0.750338 [TRT] Tactic: 786431 Time: 0.969844 [TRT] Tactic: 851967 Time: 1.27393 [TRT] Tactic: 1179647 Time: 1.12745 [TRT] Tactic: 1310719 Time: 1.99474 [TRT] Tactic: 1376255 Time: 0.759323 [TRT] Tactic: 1441791 Time: 1.28161 [TRT] Tactic: 1507327 Time: 1.25914 [TRT] Tactic: 1638399 Time: 1.47604 [TRT] Tactic: 1835007 Time: 1.00823 [TRT] Tactic: 1900543 Time: 1.1801 [TRT] Tactic: 2097151 Time: 1.14164 [TRT] Tactic: 2162687 Time: 0.787968 [TRT] Tactic: 2293759 Time: 0.825989 [TRT] Tactic: 2359295 Time: 0.990677 [TRT] Tactic: 2686975 Time: 0.914114 [TRT] Tactic: 3080191 Time: 0.907943 [TRT] Tactic: 3342335 Time: 1.17175 [TRT] Tactic: 3407871 Time: 0.881068 [TRT] Tactic: 3538943 Time: 0.966797 [TRT] Tactic: 3670015 Time: 0.728099 [TRT] Tactic: 3932159 Time: 1.377 [TRT] Tactic: 3997695 Time: 0.999818 [TRT] Tactic: 4063231 Time: 1.06299 [TRT] Tactic: 4194303 Time: 0.928985 [TRT] Tactic: 4259839 Time: 1.1675 [TRT] Tactic: 4325375 Time: 1.21089 [TRT] Tactic: 4521983 Time: 1.17706 [TRT] Tactic: 4587519 Time: 1.10909 [TRT] Tactic: 4653055 Time: 1.08063 [TRT] Tactic: 4915199 Time: 0.937083 [TRT] Tactic: 4980735 Time: 1.23899 [TRT] Tactic: 5177343 Time: 1.34654 [TRT] Tactic: 5242879 Time: 0.822708 [TRT] Tactic: 5373951 Time: 1.26872 [TRT] Tactic: 5439487 Time: 1.01063 [TRT] Tactic: 5570559 Time: 0.844896 [TRT] Tactic: 5636095 Time: 1.06174 [TRT] Tactic: 5701631 Time: 0.998151 [TRT] Tactic: 5767167 Time: 1.54307 [TRT] Tactic: 5832703 Time: 0.882344 [TRT] Tactic: 5898239 Time: 0.796068 [TRT] Tactic: 6029311 Time: 0.774011 [TRT] Tactic: 6225919 Time: 0.882682 [TRT] Tactic: 6291455 Time: 1.1282 [TRT] Tactic: 6422527 Time: 0.920599 [TRT] Tactic: 6750207 Time: 0.950652 [TRT] Tactic: 6815743 Time: 0.881667 [TRT] Tactic: 6946815 Time: 1.27583 [TRT] Tactic: 7012351 Time: 1.14117 [TRT] Tactic: 7077887 Time: 0.963099 [TRT] Tactic: 7143423 Time: 1.44182 [TRT] Tactic: 7208959 Time: 1.0288 [TRT] Tactic: 7340031 Time: 0.854166 [TRT] Tactic: 7405567 Time: 0.975677 [TRT] Tactic: 7536639 Time: 1.02409 [TRT] Tactic: 7602175 Time: 1.19169 [TRT] Tactic: 7733247 Time: 0.858073 [TRT] Tactic: 7798783 Time: 0.970131 [TRT] Tactic: 8191999 Time: 1.48026 [TRT] Tactic: 8257535 Time: 0.961042 [TRT] Tactic: 8323071 Time: 0.878698 [TRT] Tactic: 8650751 Time: 1.40977 [TRT] Tactic: 8716287 Time: 0.94638 [TRT] Tactic: 9109503 Time: 1.23083 [TRT] Tactic: 9568255 Time: 0.933203 [TRT] Tactic: 9895935 Time: 0.929792 [TRT] Tactic: 10223615 Time: 0.919505 [TRT] Tactic: 10354687 Time: 1.21938 [TRT] Tactic: 10551295 Time: 0.859115 [TRT] Tactic: 10747903 Time: 0.809818 [TRT] Tactic: 10944511 Time: 1.2351 [TRT] Fastest Tactic: 3670015 Time: 0.728099 [TRT] --------------- Timing Runner: inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce (CudnnConvolution) [TRT] Tactic: 0 Time: 0.939766 [TRT] Tactic: 1 Time: 0.794063 [TRT] Tactic: 2 Time: 1.03638 [TRT] Tactic: 4 skipped. Scratch requested: 350961664, available: 33554432 [TRT] Tactic: 5 Time: 10.7269 [TRT] Fastest Tactic: 1 Time: 0.794063 [TRT] --------------- Timing Runner: inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce (CaskConvolution) [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.601953 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v0 Tactic: 1698681053543049347 [TRT] Tactic: 1698681053543049347 Time: 0.558854 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.535781 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.454271 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.536198 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v1 Tactic: 5326823351883942011 [TRT] Tactic: 5326823351883942011 Time: 0.516042 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v0 Tactic: 5500448035057547314 [TRT] Tactic: 5500448035057547314 Time: 0.509896 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.463281 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.599193 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.541588 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v1 Tactic: -6576203419454146580 [TRT] Tactic: -6576203419454146580 Time: 0.528438 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.613646 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.55836 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.538854 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.524609 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.525964 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v0 Tactic: -377491875521947884 [TRT] Tactic: -377491875521947884 Time: 0.527864 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v1 Tactic: -37215280111360163 [TRT] Tactic: -37215280111360163 Time: 0.45198 [TRT] Fastest Tactic: -37215280111360163 Time: 0.45198 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -37215280111360163 [TRT] *************** Autotuning format combination: Float(100352,1,7168,512) -> Float(58016,1,4144,296) *************** [TRT] --------------- Timing Runner: inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce (CaskConvolution) [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 3886731678879822788 [TRT] Tactic: 3886731678879822788 Time: 0.454661 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 6629944304117643200 [TRT] Tactic: 6629944304117643200 Time: 0.745105 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.751146 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.455599 [TRT] Fastest Tactic: 3886731678879822788 Time: 0.454661 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 3886731678879822788 [TRT] *************** Autotuning format combination: Half(100352,196,14,1) -> Half(58016,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce (CudnnConvolution) [TRT] Tactic: 0 Time: 1.03484 [TRT] Tactic: 1 Time: 0.825938 [TRT] Tactic: 2 Time: 1.02031 [TRT] Tactic: 4 skipped. Scratch requested: 350961664, available: 33554432 [TRT] Tactic: 5 Time: 11.0522 [TRT] Fastest Tactic: 1 Time: 0.825938 [TRT] --------------- Timing Runner: inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 1 [TRT] *************** Autotuning format combination: Half(50176,196:2,14,1) -> Half(58016,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] *************** Autotuning format combination: Half(50176,196:2,14,1) -> Half(29008,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.475755 [TRT] Tactic: 655359 Time: 0.497318 [TRT] Tactic: 786431 Time: 0.625859 [TRT] Tactic: 851967 Time: 0.673308 [TRT] Tactic: 1179647 Time: 0.467005 [TRT] Tactic: 1310719 Time: 0.990964 [TRT] Tactic: 1376255 Time: 0.374974 [TRT] Tactic: 1441791 Time: 0.62401 [TRT] Tactic: 1507327 Time: 0.676094 [TRT] Tactic: 1638399 Time: 0.766068 [TRT] Tactic: 1835007 Time: 0.644115 [TRT] Tactic: 1900543 Time: 0.616953 [TRT] Tactic: 2097151 Time: 0.764714 [TRT] Tactic: 2162687 Time: 0.395078 [TRT] Tactic: 2293759 Time: 0.419193 [TRT] Tactic: 2359295 Time: 0.504766 [TRT] Tactic: 2686975 Time: 0.70823 [TRT] Tactic: 3080191 Time: 0.509505 [TRT] Tactic: 3342335 Time: 0.627032 [TRT] Tactic: 3407871 Time: 0.425678 [TRT] Tactic: 3538943 Time: 0.461276 [TRT] Tactic: 3670015 Time: 0.475052 [TRT] Tactic: 3932159 Time: 0.618203 [TRT] Tactic: 3997695 Time: 0.626563 [TRT] Tactic: 4063231 Time: 0.575 [TRT] Tactic: 4194303 Time: 0.540026 [TRT] Tactic: 4259839 Time: 0.758515 [TRT] Tactic: 4325375 Time: 0.623229 [TRT] Tactic: 4521983 Time: 0.633307 [TRT] Tactic: 4587519 Time: 0.651119 [TRT] Tactic: 4653055 Time: 0.557057 [TRT] Tactic: 4915199 Time: 0.543073 [TRT] Tactic: 4980735 Time: 0.629843 [TRT] Tactic: 5177343 Time: 0.573255 [TRT] Tactic: 5242879 Time: 0.38487 [TRT] Tactic: 5373951 Time: 0.560208 [TRT] Tactic: 5439487 Time: 0.567578 [TRT] Tactic: 5570559 Time: 0.548177 [TRT] Tactic: 5636095 Time: 0.573438 [TRT] Tactic: 5701631 Time: 0.451484 [TRT] Tactic: 5767167 Time: 0.709324 [TRT] Tactic: 5832703 Time: 0.41625 [TRT] Tactic: 5898239 Time: 0.46685 [TRT] Tactic: 6029311 Time: 0.371172 [TRT] Tactic: 6225919 Time: 0.406719 [TRT] Tactic: 6291455 Time: 0.468047 [TRT] Tactic: 6422527 Time: 0.463099 [TRT] Tactic: 6750207 Time: 0.549141 [TRT] Tactic: 6815743 Time: 0.421719 [TRT] Tactic: 6946815 Time: 0.639454 [TRT] Tactic: 7012351 Time: 0.762187 [TRT] Tactic: 7077887 Time: 0.440287 [TRT] Tactic: 7143423 Time: 0.722813 [TRT] Tactic: 7208959 Time: 0.492839 [TRT] Tactic: 7340031 Time: 0.508776 [TRT] Tactic: 7405567 Time: 0.478646 [TRT] Tactic: 7536639 Time: 0.499297 [TRT] Tactic: 7602175 Time: 0.581511 [TRT] Tactic: 7733247 Time: 0.473646 [TRT] Tactic: 7798783 Time: 0.627448 [TRT] Tactic: 8191999 Time: 0.75625 [TRT] Tactic: 8257535 Time: 0.563255 [TRT] Tactic: 8323071 Time: 0.510677 [TRT] Tactic: 8650751 Time: 0.683906 [TRT] Tactic: 8716287 Time: 0.462786 [TRT] Tactic: 9109503 Time: 0.77961 [TRT] Tactic: 9568255 Time: 0.54073 [TRT] Tactic: 9895935 Time: 0.537239 [TRT] Tactic: 10223615 Time: 0.707083 [TRT] Tactic: 10354687 Time: 0.766068 [TRT] Tactic: 10551295 Time: 0.433672 [TRT] Tactic: 10747903 Time: 0.417084 [TRT] Tactic: 10944511 Time: 0.630782 [TRT] Fastest Tactic: 6029311 Time: 0.371172 [TRT] --------------- Timing Runner: inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce (CaskConvolution) [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_interior_nn_v1 Tactic: 3066127711859985668 [TRT] Tactic: 3066127711859985668 Time: 0.261406 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.307526 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.275781 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.242422 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] Tactic: 8163473458334948789 Time: 0.231615 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.277449 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.281797 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.242578 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_interior_nn_v1 Tactic: -1716393687483585322 [TRT] Tactic: -1716393687483585322 Time: 0.27125 [TRT] Fastest Tactic: 8163473458334948789 Time: 0.231615 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 8163473458334948789 [TRT] *************** Autotuning Reformat:Float(58016,196,14,1) -> Float(58016,1,4144,296) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.087969 [TRT] Tactic: 0 Time: 0.094401 [TRT] Fastest Tactic: 1002 Time: 0.087969 [TRT] *************** Autotuning Reformat:Float(58016,196,14,1) -> Half(58016,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.56292 [TRT] Tactic: 0 Time: 0.0678645 [TRT] Fastest Tactic: 0 Time: 0.0678645 [TRT] *************** Autotuning Reformat:Float(58016,196,14,1) -> Half(29008,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.098984 [TRT] Tactic: 0 Time: 0.051927 [TRT] Fastest Tactic: 0 Time: 0.051927 [TRT] *************** Autotuning Reformat:Float(58016,1,4144,296) -> Float(58016,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.10888 [TRT] Tactic: 0 Time: 0.084844 [TRT] Fastest Tactic: 0 Time: 0.084844 [TRT] *************** Autotuning Reformat:Float(58016,1,4144,296) -> Half(58016,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0753125 [TRT] Tactic: 0 Time: 0.0855205 [TRT] Fastest Tactic: 1002 Time: 0.0753125 [TRT] *************** Autotuning Reformat:Float(58016,1,4144,296) -> Half(29008,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0877605 [TRT] Tactic: 0 Time: 0.101197 [TRT] Fastest Tactic: 1002 Time: 0.0877605 [TRT] *************** Autotuning Reformat:Half(58016,196,14,1) -> Float(58016,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.61172 [TRT] Tactic: 0 Time: 0.0558075 [TRT] Fastest Tactic: 0 Time: 0.0558075 [TRT] *************** Autotuning Reformat:Half(58016,196,14,1) -> Float(58016,1,4144,296) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0702605 [TRT] Tactic: 0 Time: 0.092526 [TRT] Fastest Tactic: 1002 Time: 0.0702605 [TRT] *************** Autotuning Reformat:Half(58016,196,14,1) -> Half(29008,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.070026 [TRT] Tactic: 0 Time: 0.051849 [TRT] Fastest Tactic: 0 Time: 0.051849 [TRT] *************** Autotuning Reformat:Half(29008,196:2,14,1) -> Float(58016,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0941925 [TRT] Tactic: 0 Time: 0.0453905 [TRT] Fastest Tactic: 0 Time: 0.0453905 [TRT] *************** Autotuning Reformat:Half(29008,196:2,14,1) -> Float(58016,1,4144,296) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0703645 [TRT] Tactic: 0 Time: 0.107239 [TRT] Fastest Tactic: 1002 Time: 0.0703645 [TRT] *************** Autotuning Reformat:Half(29008,196:2,14,1) -> Half(58016,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.128541 [TRT] Tactic: 0 Time: 0.0432555 [TRT] Fastest Tactic: 0 Time: 0.0432555 [TRT] *************** Autotuning Reformat:Float(58016,196,14,1) -> Float(58016,1,4144,296) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0371355 [TRT] Tactic: 0 Time: 0.0370055 [TRT] Fastest Tactic: 0 Time: 0.0370055 [TRT] *************** Autotuning Reformat:Float(58016,196,14,1) -> Half(58016,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.718984 [TRT] Tactic: 0 Time: 0.033463 [TRT] Fastest Tactic: 0 Time: 0.033463 [TRT] *************** Autotuning Reformat:Float(58016,196,14,1) -> Half(29008,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.04276 [TRT] Tactic: 0 Time: 0.0403385 [TRT] Fastest Tactic: 0 Time: 0.0403385 [TRT] *************** Autotuning Reformat:Float(58016,1,4144,296) -> Float(58016,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.045339 [TRT] Tactic: 0 Time: 0.033672 [TRT] Fastest Tactic: 0 Time: 0.033672 [TRT] *************** Autotuning Reformat:Float(58016,1,4144,296) -> Half(58016,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.033542 [TRT] Tactic: 0 Time: 0.0356775 [TRT] Fastest Tactic: 1002 Time: 0.033542 [TRT] *************** Autotuning Reformat:Float(58016,1,4144,296) -> Half(29008,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.036849 [TRT] Tactic: 0 Time: 0.040417 [TRT] Fastest Tactic: 1002 Time: 0.036849 [TRT] *************** Autotuning Reformat:Half(58016,196,14,1) -> Float(58016,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.73539 [TRT] Tactic: 0 Time: 0.021901 [TRT] Fastest Tactic: 0 Time: 0.021901 [TRT] *************** Autotuning Reformat:Half(58016,196,14,1) -> Float(58016,1,4144,296) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0311715 [TRT] Tactic: 0 Time: 0.038151 [TRT] Fastest Tactic: 1002 Time: 0.0311715 [TRT] *************** Autotuning Reformat:Half(58016,196,14,1) -> Half(29008,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0314055 [TRT] Tactic: 0 Time: 0.021823 [TRT] Fastest Tactic: 0 Time: 0.021823 [TRT] *************** Autotuning Reformat:Half(29008,196:2,14,1) -> Float(58016,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0404685 [TRT] Tactic: 0 Time: 0.019505 [TRT] Fastest Tactic: 0 Time: 0.019505 [TRT] *************** Autotuning Reformat:Half(29008,196:2,14,1) -> Float(58016,1,4144,296) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.031198 [TRT] Tactic: 0 Time: 0.042734 [TRT] Fastest Tactic: 1002 Time: 0.031198 [TRT] *************** Autotuning Reformat:Half(29008,196:2,14,1) -> Half(58016,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.052031 [TRT] Tactic: 0 Time: 0.035651 [TRT] Fastest Tactic: 0 Time: 0.035651 [TRT] *************** Autotuning format combination: Float(58016,196,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/3x3 + inception_4b/relu_3x3 (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4b/3x3 + inception_4b/relu_3x3 (FusedConvActConvolution) [TRT] Tactic: 524287 Time: 0.644506 [TRT] Tactic: 720895 Time: 0.78862 [TRT] Tactic: 983039 Time: 0.673568 [TRT] Tactic: 1048575 Time: 0.655755 [TRT] Tactic: 1703935 Time: 0.641641 [TRT] Tactic: 1769471 Time: 0.70776 [TRT] Tactic: 1966079 Time: 0.749531 [TRT] Tactic: 2031615 Time: 0.721328 [TRT] Tactic: 2228223 Time: 0.834454 [TRT] Tactic: 2424831 Time: 0.907474 [TRT] Tactic: 2621439 Time: 0.774714 [TRT] Tactic: 2752511 Time: 0.642031 [TRT] Tactic: 2818047 Time: 0.90461 [TRT] Tactic: 2883583 Time: 0.706015 [TRT] Tactic: 3014655 Time: 0.601068 [TRT] Tactic: 3145727 Time: 0.638568 [TRT] Tactic: 3473407 Time: 0.791719 [TRT] Tactic: 3604479 Time: 0.593463 [TRT] Tactic: 3735551 Time: 0.871432 [TRT] Tactic: 4390911 Time: 0.653021 [TRT] Tactic: 5046271 Time: 0.628229 [TRT] Tactic: 5963775 Time: 0.684739 [TRT] Tactic: 6160383 Time: 0.626824 [TRT] Tactic: 6488063 Time: 0.561823 [TRT] Tactic: 6881279 Time: 0.670833 [TRT] Tactic: 7274495 Time: 0.88526 [TRT] Tactic: 7864319 Time: 0.807865 [TRT] Tactic: 7995391 Time: 0.766719 [TRT] Tactic: 8585215 Time: 0.5975 [TRT] Tactic: 8847359 Time: 0.607474 [TRT] Tactic: 8978431 Time: 0.690547 [TRT] Tactic: 9043967 Time: 0.548854 [TRT] Tactic: 9175039 Time: 0.593412 [TRT] Tactic: 9502719 Time: 0.678125 [TRT] Tactic: 9830399 Time: 0.667005 [TRT] Tactic: 9961471 Time: 0.675625 [TRT] Tactic: 10027007 Time: 0.569219 [TRT] Tactic: 10092543 Time: 0.653333 [TRT] Tactic: 10289151 Time: 0.749974 [TRT] Tactic: 10485759 Time: 0.564322 [TRT] Tactic: 10682367 Time: 0.772135 [TRT] Tactic: 10813439 Time: 0.777552 [TRT] Fastest Tactic: 9043967 Time: 0.548854 [TRT] --------------- Timing Runner: inception_4b/3x3 + inception_4b/relu_3x3 (CudnnConvolution) [TRT] Tactic: 0 Time: 1.32326 [TRT] Tactic: 1 Time: 0.812318 [TRT] Tactic: 2 Time: 1.22287 [TRT] Tactic: 4 skipped. Scratch requested: 58963968, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 110645248, available: 33554432 [TRT] Tactic: 6 Time: 0.590834 [TRT] Fastest Tactic: 6 Time: 0.590834 [TRT] --------------- Timing Runner: inception_4b/3x3 + inception_4b/relu_3x3 (CaskConvolution) [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.847031 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v0 Tactic: 1754984623894446479 [TRT] Tactic: 1754984623894446479 Time: 0.899219 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v0 Tactic: 3611739942397549984 [TRT] Tactic: 3611739942397549984 Time: 0.692421 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148n_nt_v1 Tactic: 3827454225649558724 [TRT] Tactic: 3827454225649558724 Time: 0.520938 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v1 Tactic: 4337000649858996379 [TRT] Tactic: 4337000649858996379 Time: 0.716251 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.694219 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.673203 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.678177 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 5921334924264294896 [TRT] Tactic: 5921334924264294896 Time: 0.392162 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.697865 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.768541 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148t_nt_v1 Tactic: 7852627285308570038 [TRT] Tactic: 7852627285308570038 Time: 0.505834 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v1 Tactic: -9137461792520977713 [TRT] Tactic: -9137461792520977713 Time: 0.698178 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148n_nt_v0 Tactic: -8776506421218919509 [TRT] Tactic: -8776506421218919509 Time: 0.494635 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.697968 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v0 Tactic: -8133971918129952780 [TRT] Tactic: -8133971918129952780 Time: 0.746901 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v1 Tactic: -6092040395344634144 [TRT] Tactic: -6092040395344634144 Time: 0.892812 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.890208 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.734766 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148t_nt_v0 Tactic: -2318106587342035239 [TRT] Tactic: -2318106587342035239 Time: 0.486198 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_mobile_relu_tile148t_nt_v0 Tactic: -1343271414618805657 [TRT] Tactic: -1343271414618805657 Time: 0.355834 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.74112 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.712344 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.667318 [TRT] Fastest Tactic: -1343271414618805657 Time: 0.355834 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -1343271414618805657 [TRT] *************** Autotuning format combination: Float(58016,1,4144,296) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: inception_4b/3x3 + inception_4b/relu_3x3 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4b/3x3 + inception_4b/relu_3x3 (CaskConvolution) [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.854297 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.662421 [TRT] Fastest Tactic: -7394439838318485025 Time: 0.662421 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(58016,196,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/3x3 + inception_4b/relu_3x3 (CudnnConvolution) [TRT] Tactic: 0 Time: 1.36042 [TRT] Tactic: 1 Time: 1.3638 [TRT] Tactic: 2 Time: 1.19568 [TRT] Tactic: 4 skipped. Scratch requested: 58963968, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 110645248, available: 33554432 [TRT] Tactic: 6 Time: 1.8119 [TRT] Fastest Tactic: 2 Time: 1.19568 [TRT] --------------- Timing Runner: inception_4b/3x3 + inception_4b/relu_3x3 (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 2 [TRT] *************** Autotuning format combination: Half(29008,196:2,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/3x3 + inception_4b/relu_3x3 (FusedConvActConvolution) [TRT] Tactic: 524287 Time: 0.340026 [TRT] Tactic: 720895 Time: 0.474245 [TRT] Tactic: 983039 Time: 0.37776 [TRT] Tactic: 1048575 Time: 0.369141 [TRT] Tactic: 1703935 Time: 0.372474 [TRT] Tactic: 1769471 Time: 2.35034 [TRT] Tactic: 1966079 Time: 0.43888 [TRT] Tactic: 2031615 Time: 0.377213 [TRT] Tactic: 2228223 Time: 0.46711 [TRT] Tactic: 2621439 Time: 0.434192 [TRT] Tactic: 2752511 Time: 0.382812 [TRT] Tactic: 2818047 Time: 0.498568 [TRT] Tactic: 2883583 Time: 0.454271 [TRT] Tactic: 3014655 Time: 0.353307 [TRT] Tactic: 3145727 Time: 0.383125 [TRT] Tactic: 3473407 Time: 0.465026 [TRT] Tactic: 3604479 Time: 0.349635 [TRT] Tactic: 3735551 Time: 0.414818 [TRT] Tactic: 4390911 Time: 0.381666 [TRT] Tactic: 5046271 Time: 0.344375 [TRT] Tactic: 5963775 Time: 0.365703 [TRT] Tactic: 6160383 Time: 0.35987 [TRT] Tactic: 6488063 Time: 0.321302 [TRT] Tactic: 6881279 Time: 0.348099 [TRT] Tactic: 7274495 Time: 0.493464 [TRT] Tactic: 7864319 Time: 0.465599 [TRT] Tactic: 7995391 Time: 0.468073 [TRT] Tactic: 8585215 Time: 0.335286 [TRT] Tactic: 8847359 Time: 0.34776 [TRT] Tactic: 8978431 Time: 0.388386 [TRT] Tactic: 9043967 Time: 0.312057 [TRT] Tactic: 9175039 Time: 0.350625 [TRT] Tactic: 9502719 Time: 0.375912 [TRT] Tactic: 9830399 Time: 0.355469 [TRT] Tactic: 10027007 Time: 0.31776 [TRT] Tactic: 10092543 Time: 0.375781 [TRT] Tactic: 10289151 Time: 0.437969 [TRT] Tactic: 10485759 Time: 0.315677 [TRT] Tactic: 10682367 Time: 0.426641 [TRT] Tactic: 10813439 Time: 0.472214 [TRT] Fastest Tactic: 9043967 Time: 0.312057 [TRT] --------------- Timing Runner: inception_4b/3x3 + inception_4b/relu_3x3 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4b/3x3 + inception_4b/relu_3x3 (CaskConvolution) [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.437604 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_large_nn_v1 Tactic: 3650389455493082349 [TRT] Tactic: 3650389455493082349 Time: 0.466927 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_winograd_fp16x2_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 4772821744921268633 [TRT] Tactic: 4772821744921268633 Time: 0.211875 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.372604 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.36375 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_large_nn_v1 Tactic: -6490690591794140522 [TRT] Tactic: -6490690591794140522 Time: 0.369922 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_large_nn_v1 Tactic: -4686027666808657977 [TRT] Tactic: -4686027666808657977 Time: 0.358229 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.341223 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.353333 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.349167 [TRT] Fastest Tactic: 4772821744921268633 Time: 0.211875 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 4772821744921268633 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0728125 [TRT] Tactic: 0 Time: 0.071198 [TRT] Fastest Tactic: 0 Time: 0.071198 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.43052 [TRT] Tactic: 0 Time: 0.065156 [TRT] Fastest Tactic: 0 Time: 0.065156 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.075313 [TRT] Tactic: 0 Time: 0.040573 [TRT] Fastest Tactic: 0 Time: 0.040573 [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.088281 [TRT] Tactic: 0 Time: 0.067083 [TRT] Fastest Tactic: 0 Time: 0.067083 [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0550525 [TRT] Tactic: 0 Time: 0.065625 [TRT] Fastest Tactic: 1002 Time: 0.0550525 [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.065391 [TRT] Tactic: 0 Time: 0.0768495 [TRT] Fastest Tactic: 1002 Time: 0.065391 [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.4624 [TRT] Tactic: 0 Time: 0.065859 [TRT] Fastest Tactic: 0 Time: 0.065859 [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.05 [TRT] Tactic: 0 Time: 0.0701825 [TRT] Fastest Tactic: 1002 Time: 0.05 [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.054557 [TRT] Tactic: 0 Time: 0.0403645 [TRT] Fastest Tactic: 0 Time: 0.0403645 [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.07151 [TRT] Tactic: 0 Time: 0.076016 [TRT] Fastest Tactic: 1002 Time: 0.07151 [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0504425 [TRT] Tactic: 0 Time: 0.081979 [TRT] Fastest Tactic: 1002 Time: 0.0504425 [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.096094 [TRT] Tactic: 0 Time: 0.068255 [TRT] Fastest Tactic: 0 Time: 0.068255 [TRT] *************** Autotuning Reformat:Float(58016,196,14,1) -> Float(58016,1,4144,296) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.024141 [TRT] Tactic: 0 Time: 0.0101565 [TRT] Fastest Tactic: 0 Time: 0.0101565 [TRT] *************** Autotuning Reformat:Float(58016,196,14,1) -> Half(58016,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.158724 [TRT] Tactic: 0 Time: 0.0102085 [TRT] Fastest Tactic: 0 Time: 0.0102085 [TRT] *************** Autotuning Reformat:Float(58016,196,14,1) -> Half(29008,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.01276 [TRT] Tactic: 0 Time: 0.012526 [TRT] Fastest Tactic: 0 Time: 0.012526 [TRT] *************** Autotuning Reformat:Float(58016,1,4144,296) -> Float(58016,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.026485 [TRT] Tactic: 0 Time: 0.0104165 [TRT] Fastest Tactic: 0 Time: 0.0104165 [TRT] *************** Autotuning Reformat:Float(58016,1,4144,296) -> Half(58016,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.01263 [TRT] Tactic: 0 Time: 0.0104165 [TRT] Fastest Tactic: 0 Time: 0.0104165 [TRT] *************** Autotuning Reformat:Float(58016,1,4144,296) -> Half(29008,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.012735 [TRT] Tactic: 0 Time: 0.0126045 [TRT] Fastest Tactic: 0 Time: 0.0126045 [TRT] *************** Autotuning Reformat:Half(58016,196,14,1) -> Float(58016,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.163151 [TRT] Tactic: 0 Time: 0.0079425 [TRT] Fastest Tactic: 0 Time: 0.0079425 [TRT] *************** Autotuning Reformat:Half(58016,196,14,1) -> Float(58016,1,4144,296) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0119015 [TRT] Tactic: 0 Time: 0.0102345 [TRT] Fastest Tactic: 0 Time: 0.0102345 [TRT] *************** Autotuning Reformat:Half(58016,196,14,1) -> Half(29008,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.012656 [TRT] Tactic: 0 Time: 0.0078125 [TRT] Fastest Tactic: 0 Time: 0.0078125 [TRT] *************** Autotuning Reformat:Half(29008,196:2,14,1) -> Float(58016,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0148955 [TRT] Tactic: 0 Time: 0.007812 [TRT] Fastest Tactic: 0 Time: 0.007812 [TRT] *************** Autotuning Reformat:Half(29008,196:2,14,1) -> Float(58016,1,4144,296) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.012578 [TRT] Tactic: 0 Time: 0.012552 [TRT] Fastest Tactic: 0 Time: 0.012552 [TRT] *************** Autotuning Reformat:Half(29008,196:2,14,1) -> Half(58016,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0241405 [TRT] Tactic: 0 Time: 0.01026 [TRT] Fastest Tactic: 0 Time: 0.01026 [TRT] *************** Autotuning format combination: Float(58016,196,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/5x5 + inception_4b/relu_5x5 (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4b/5x5 + inception_4b/relu_5x5 (FusedConvActConvolution) [TRT] Tactic: 393215 Time: 0.137786 [TRT] Tactic: 917503 Time: 0.12487 [TRT] Tactic: 1114111 Time: 0.12927 [TRT] Tactic: 1245183 Time: 0.132135 [TRT] Tactic: 1572863 Time: 0.138386 [TRT] Tactic: 2490367 Time: 0.150833 [TRT] Tactic: 2555903 Time: 0.140703 [TRT] Tactic: 2949119 Time: 0.115573 [TRT] Tactic: 3211263 Time: 0.226432 [TRT] Tactic: 3801087 Time: 0.128881 [TRT] Tactic: 3866623 Time: 0.130443 [TRT] Tactic: 4128767 Time: 0.114479 [TRT] Tactic: 4456447 Time: 0.119817 [TRT] Tactic: 4718591 Time: 0.115964 [TRT] Tactic: 4784127 Time: 0.294324 [TRT] Tactic: 4849663 Time: 0.121355 [TRT] Tactic: 5111807 Time: 0.119896 [TRT] Tactic: 5308415 Time: 0.138255 [TRT] Tactic: 5505023 Time: 0.200938 [TRT] Tactic: 6094847 Time: 0.13375 [TRT] Tactic: 6356991 Time: 0.141041 [TRT] Tactic: 6553599 Time: 0.139037 [TRT] Tactic: 6619135 Time: 0.150703 [TRT] Tactic: 6684671 Time: 0.213229 [TRT] Tactic: 7471103 Time: 0.131172 [TRT] Tactic: 7667711 Time: 0.116328 [TRT] Tactic: 7929855 Time: 0.12776 [TRT] Tactic: 8060927 Time: 0.142343 [TRT] Tactic: 8126463 Time: 0.124506 [TRT] Tactic: 8388607 Time: 0.137942 [TRT] Tactic: 8519679 Time: 0.156276 [TRT] Tactic: 8781823 Time: 0.144922 [TRT] Tactic: 8912895 Time: 0.124427 [TRT] Tactic: 9240575 Time: 0.114558 [TRT] Tactic: 9306111 Time: 0.1925 [TRT] Tactic: 9371647 Time: 0.125547 [TRT] Tactic: 9437183 Time: 0.114844 [TRT] Tactic: 9633791 Time: 0.120026 [TRT] Tactic: 9699327 Time: 0.137708 [TRT] Tactic: 9764863 Time: 0.143854 [TRT] Tactic: 10158079 Time: 0.14237 [TRT] Tactic: 10420223 Time: 0.129688 [TRT] Tactic: 10616831 Time: 0.12776 [TRT] Tactic: 10878975 Time: 0.150078 [TRT] Fastest Tactic: 4128767 Time: 0.114479 [TRT] --------------- Timing Runner: inception_4b/5x5 + inception_4b/relu_5x5 (CudnnConvolution) [TRT] Tactic: 0 Time: 0.264531 [TRT] Tactic: 1 Time: 0.277709 [TRT] Tactic: 2 Time: 0.437499 [TRT] Tactic: 4 Time: 2.80346 [TRT] Tactic: 5 Time: 2.58102 [TRT] Fastest Tactic: 0 Time: 0.264531 [TRT] --------------- Timing Runner: inception_4b/5x5 + inception_4b/relu_5x5 (CaskConvolution) [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.189323 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v0 Tactic: 1754984623894446479 [TRT] Tactic: 1754984623894446479 Time: 0.186042 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v0 Tactic: 3611739942397549984 [TRT] Tactic: 3611739942397549984 Time: 0.223203 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v1 Tactic: 4337000649858996379 [TRT] Tactic: 4337000649858996379 Time: 0.15112 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.221224 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.121771 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.210938 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.145104 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.139323 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v1 Tactic: -9137461792520977713 [TRT] Tactic: -9137461792520977713 Time: 0.226406 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.226901 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v0 Tactic: -8133971918129952780 [TRT] Tactic: -8133971918129952780 Time: 0.160312 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v1 Tactic: -6092040395344634144 [TRT] Tactic: -6092040395344634144 Time: 0.200078 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.18461 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.133593 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.158802 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.138204 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.209583 [TRT] Fastest Tactic: 5137655947464784826 Time: 0.121771 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 4128767 [TRT] *************** Autotuning format combination: Float(58016,1,4144,296) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: inception_4b/5x5 + inception_4b/relu_5x5 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4b/5x5 + inception_4b/relu_5x5 (CaskConvolution) [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.186406 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.142969 [TRT] Fastest Tactic: -7394439838318485025 Time: 0.142969 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(58016,196,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/5x5 + inception_4b/relu_5x5 (CudnnConvolution) [TRT] Tactic: 0 Time: 0.282552 [TRT] Tactic: 1 Time: 0.281927 [TRT] Tactic: 2 Time: 0.433255 [TRT] Tactic: 4 Time: 2.81031 [TRT] Tactic: 5 Time: 2.58117 [TRT] Fastest Tactic: 1 Time: 0.281927 [TRT] --------------- Timing Runner: inception_4b/5x5 + inception_4b/relu_5x5 (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 1 [TRT] *************** Autotuning format combination: Half(29008,196:2,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/5x5 + inception_4b/relu_5x5 (FusedConvActConvolution) [TRT] Tactic: 393215 Time: 0.0786455 [TRT] Tactic: 917503 Time: 0.075599 [TRT] Tactic: 1114111 Time: 0.0692185 [TRT] Tactic: 1245183 Time: 0.0728645 [TRT] Tactic: 1572863 Time: 0.0802605 [TRT] Tactic: 3211263 Time: 0.134297 [TRT] Tactic: 3801087 Time: 0.0732555 [TRT] Tactic: 3866623 Time: 0.0744015 [TRT] Tactic: 4128767 Time: 0.0650785 [TRT] Tactic: 4456447 Time: 0.0721615 [TRT] Tactic: 4718591 Time: 0.0665105 [TRT] Tactic: 4784127 Time: 0.151719 [TRT] Tactic: 4849663 Time: 0.067682 [TRT] Tactic: 5111807 Time: 0.068125 [TRT] Tactic: 5308415 Time: 0.080026 [TRT] Tactic: 6094847 Time: 0.075651 [TRT] Tactic: 6553599 Time: 0.077891 [TRT] Tactic: 6619135 Time: 0.0800775 [TRT] Tactic: 7471103 Time: 0.076901 [TRT] Tactic: 7667711 Time: 0.0651565 [TRT] Tactic: 7929855 Time: 0.0683855 [TRT] Tactic: 8781823 Time: 0.0763805 [TRT] Tactic: 9240575 Time: 0.0617965 [TRT] Tactic: 9306111 Time: 0.099531 [TRT] Tactic: 9371647 Time: 0.067734 [TRT] Tactic: 9633791 Time: 0.0660935 [TRT] Tactic: 9699327 Time: 0.079479 [TRT] Tactic: 9764863 Time: 0.0879425 [TRT] Tactic: 10158079 Time: 0.0819795 [TRT] Tactic: 10616831 Time: 0.073932 [TRT] Tactic: 10878975 Time: 0.088307 [TRT] Fastest Tactic: 9240575 Time: 0.0617965 [TRT] --------------- Timing Runner: inception_4b/5x5 + inception_4b/relu_5x5 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4b/5x5 + inception_4b/relu_5x5 (CaskConvolution) [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.10526 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_large_nn_v1 Tactic: 3650389455493082349 [TRT] Tactic: 3650389455493082349 Time: 0.108021 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.0753125 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.0823965 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_large_nn_v1 Tactic: -6490690591794140522 [TRT] Tactic: -6490690591794140522 Time: 0.084896 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_large_nn_v1 Tactic: -4686027666808657977 [TRT] Tactic: -4686027666808657977 Time: 0.122187 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.111171 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.120547 [TRT] inception_4b/5x5 + inception_4b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.066068 [TRT] Fastest Tactic: -2409163523992614473 Time: 0.066068 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 9240575 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning format combination: Float(100352,196,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/pool (TiledPooling) [TRT] Tactic: 2752769 Time: 0.215807 [TRT] Tactic: 2818305 Time: 0.20474 [TRT] Tactic: 2883841 Time: 0.152396 [TRT] Tactic: 2949377 Time: 0.825598 [TRT] Tactic: 3014913 Time: 0.743073 [TRT] Tactic: 3080449 Time: 0.408985 [TRT] Tactic: 3145985 Time: 0.331198 [TRT] Tactic: 3211521 Time: 0.139037 [TRT] Tactic: 3277057 Time: 0.135 [TRT] Tactic: 3342593 Time: 0.105547 [TRT] Tactic: 3408129 Time: 0.478386 [TRT] Tactic: 3473665 Time: 0.432136 [TRT] Tactic: 3539201 Time: 0.252474 [TRT] Tactic: 3604737 Time: 0.201042 [TRT] Tactic: 3670273 Time: 0.110208 [TRT] Tactic: 3735809 Time: 0.110312 [TRT] Tactic: 3801345 Time: 0.0904425 [TRT] Tactic: 3866881 Time: 0.375885 [TRT] Tactic: 3932417 Time: 0.331823 [TRT] Tactic: 3997953 Time: 0.203698 [TRT] Tactic: 4063489 Time: 0.164427 [TRT] Tactic: 4129025 Time: 0.109062 [TRT] Tactic: 4194561 Time: 0.108255 [TRT] Tactic: 4260097 Time: 0.0851565 [TRT] Tactic: 4325633 Time: 0.321432 [TRT] Tactic: 4391169 Time: 0.283932 [TRT] Tactic: 4456705 Time: 0.179036 [TRT] Tactic: 4522241 Time: 0.147031 [TRT] Tactic: 4587777 Time: 0.109271 [TRT] Tactic: 4653313 Time: 0.109975 [TRT] Tactic: 4718849 Time: 0.0814585 [TRT] Tactic: 4784385 Time: 0.297369 [TRT] Tactic: 4849921 Time: 0.268333 [TRT] Tactic: 4915457 Time: 0.169454 [TRT] Tactic: 4980993 Time: 0.135755 [TRT] Tactic: 5046529 Time: 0.111224 [TRT] Tactic: 5112065 Time: 0.108855 [TRT] Tactic: 5177601 Time: 0.0808855 [TRT] Tactic: 5243137 Time: 0.277786 [TRT] Tactic: 5308673 Time: 0.248698 [TRT] Tactic: 5374209 Time: 0.162109 [TRT] Tactic: 5439745 Time: 0.129609 [TRT] Tactic: 6553857 Time: 0.160338 [TRT] Tactic: 6750465 Time: 0.108854 [TRT] Fastest Tactic: 5177601 Time: 0.0808855 [TRT] --------------- Timing Runner: inception_4b/pool (CudnnPooling) [TRT] Tactic: -1 Time: 0.321563 [TRT] Fastest Tactic: -1 Time: 0.321563 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: TiledPooling Tactic: 5177601 [TRT] *************** Autotuning format combination: Half(100352,196,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/pool (TiledPooling) [TRT] TiledPooling has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4b/pool (CudnnPooling) [TRT] Tactic: -1 Time: 0.335912 [TRT] Fastest Tactic: -1 Time: 0.335912 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnPooling Tactic: -1 [TRT] *************** Autotuning format combination: Half(50176,196:2,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/pool (TiledPooling) [TRT] Tactic: 2752769 Time: 0.120156 [TRT] Tactic: 2818305 Time: 0.119505 [TRT] Tactic: 2883841 Time: 0.09112 [TRT] Tactic: 2949377 Time: 0.448568 [TRT] Tactic: 3014913 Time: 0.417526 [TRT] Tactic: 3080449 Time: 0.23513 [TRT] Tactic: 3145985 Time: 0.185625 [TRT] Tactic: 3211521 Time: 0.0863805 [TRT] Tactic: 3277057 Time: 0.0851825 [TRT] Tactic: 3342593 Time: 0.0677605 [TRT] Tactic: 3408129 Time: 0.3 [TRT] Tactic: 3473665 Time: 0.272891 [TRT] Tactic: 3539201 Time: 0.157578 [TRT] Tactic: 3604737 Time: 0.131458 [TRT] Tactic: 3670273 Time: 0.0752865 [TRT] Tactic: 3735809 Time: 0.0725 [TRT] Tactic: 3801345 Time: 0.059036 [TRT] Tactic: 3866881 Time: 0.258125 [TRT] Tactic: 3932417 Time: 0.232343 [TRT] Tactic: 3997953 Time: 0.138307 [TRT] Tactic: 4063489 Time: 0.11625 [TRT] Tactic: 4129025 Time: 0.067943 [TRT] Tactic: 4194561 Time: 0.0678645 [TRT] Tactic: 4260097 Time: 0.056693 [TRT] Tactic: 4325633 Time: 0.226406 [TRT] Tactic: 4391169 Time: 0.211719 [TRT] Tactic: 4456705 Time: 0.13901 [TRT] Tactic: 4522241 Time: 0.115313 [TRT] Tactic: 4587777 Time: 0.073594 [TRT] Tactic: 4653313 Time: 0.0730205 [TRT] Tactic: 4718849 Time: 0.060052 [TRT] Tactic: 4784385 Time: 0.239349 [TRT] Tactic: 4849921 Time: 0.220755 [TRT] Tactic: 4915457 Time: 0.133802 [TRT] Tactic: 4980993 Time: 0.113255 [TRT] Tactic: 5046529 Time: 0.0703385 [TRT] Tactic: 5112065 Time: 0.070156 [TRT] Tactic: 5177601 Time: 0.0582555 [TRT] Tactic: 5243137 Time: 0.229557 [TRT] Tactic: 5308673 Time: 0.211146 [TRT] Tactic: 5374209 Time: 0.128047 [TRT] Tactic: 5439745 Time: 0.109557 [TRT] Tactic: 6553857 Time: 0.112057 [TRT] Tactic: 6750465 Time: 0.0799995 [TRT] Fastest Tactic: 4260097 Time: 0.056693 [TRT] --------------- Timing Runner: inception_4b/pool (CudaPooling) [TRT] Tactic: -3 Time: 0.197031 [TRT] Fastest Tactic: -3 Time: 0.197031 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: TiledPooling Tactic: 4260097 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning format combination: Float(100352,196,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/pool_proj + inception_4b/relu_pool_proj (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4b/pool_proj + inception_4b/relu_pool_proj (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.22401 [TRT] Tactic: 655359 Time: 0.250651 [TRT] Tactic: 786431 Time: 0.216406 [TRT] Tactic: 851967 Time: 0.359062 [TRT] Tactic: 1179647 Time: 0.199401 [TRT] Tactic: 1310719 Time: 0.452577 [TRT] Tactic: 1376255 Time: 0.234037 [TRT] Tactic: 1441791 Time: 0.252865 [TRT] Tactic: 1507327 Time: 0.473907 [TRT] Tactic: 1638399 Time: 0.299688 [TRT] Tactic: 1835007 Time: 0.227605 [TRT] Tactic: 1900543 Time: 0.45586 [TRT] Tactic: 2097151 Time: 0.242161 [TRT] Tactic: 2162687 Time: 0.25086 [TRT] Tactic: 2293759 Time: 0.236667 [TRT] Tactic: 2359295 Time: 0.261797 [TRT] Tactic: 2686975 Time: 0.233854 [TRT] Tactic: 3080191 Time: 0.285703 [TRT] Tactic: 3342335 Time: 0.375182 [TRT] Tactic: 3407871 Time: 0.180521 [TRT] Tactic: 3538943 Time: 0.179114 [TRT] Tactic: 3670015 Time: 0.288333 [TRT] Tactic: 3932159 Time: 0.406745 [TRT] Tactic: 3997695 Time: 0.220625 [TRT] Tactic: 4063231 Time: 0.314297 [TRT] Tactic: 4194303 Time: 0.217344 [TRT] Tactic: 4259839 Time: 0.252032 [TRT] Tactic: 4325375 Time: 0.280625 [TRT] Tactic: 4521983 Time: 0.273099 [TRT] Tactic: 4587519 Time: 0.244245 [TRT] Tactic: 4653055 Time: 0.218958 [TRT] Tactic: 4915199 Time: 0.227605 [TRT] Tactic: 4980735 Time: 0.335443 [TRT] Tactic: 5177343 Time: 0.232631 [TRT] Tactic: 5242879 Time: 0.161406 [TRT] Tactic: 5373951 Time: 0.226198 [TRT] Tactic: 5439487 Time: 0.215573 [TRT] Tactic: 5570559 Time: 0.244401 [TRT] Tactic: 5636095 Time: 0.314219 [TRT] Tactic: 5701631 Time: 0.236927 [TRT] Tactic: 5767167 Time: 0.301537 [TRT] Tactic: 5832703 Time: 0.176979 [TRT] Tactic: 5898239 Time: 0.162865 [TRT] Tactic: 6029311 Time: 0.228073 [TRT] Tactic: 6225919 Time: 0.158073 [TRT] Tactic: 6291455 Time: 0.201016 [TRT] Tactic: 6422527 Time: 0.302891 [TRT] Tactic: 6750207 Time: 0.214817 [TRT] Tactic: 6815743 Time: 0.163282 [TRT] Tactic: 6946815 Time: 0.304818 [TRT] Tactic: 7012351 Time: 0.242188 [TRT] Tactic: 7077887 Time: 0.173307 [TRT] Tactic: 7143423 Time: 0.315834 [TRT] Tactic: 7208959 Time: 0.247005 [TRT] Tactic: 7340031 Time: 0.17151 [TRT] Tactic: 7405567 Time: 0.199948 [TRT] Tactic: 7536639 Time: 0.187943 [TRT] Tactic: 7602175 Time: 0.292396 [TRT] Tactic: 7733247 Time: 0.167292 [TRT] Tactic: 7798783 Time: 0.215755 [TRT] Tactic: 8191999 Time: 0.324506 [TRT] Tactic: 8257535 Time: 0.236042 [TRT] Tactic: 8323071 Time: 0.207969 [TRT] Tactic: 8650751 Time: 0.303855 [TRT] Tactic: 8716287 Time: 0.165781 [TRT] Tactic: 9109503 Time: 0.258073 [TRT] Tactic: 9568255 Time: 0.226745 [TRT] Tactic: 9895935 Time: 0.218098 [TRT] Tactic: 10223615 Time: 0.235312 [TRT] Tactic: 10354687 Time: 0.265078 [TRT] Tactic: 10551295 Time: 0.194792 [TRT] Tactic: 10747903 Time: 0.167812 [TRT] Tactic: 10944511 Time: 0.3325 [TRT] Fastest Tactic: 6225919 Time: 0.158073 [TRT] --------------- Timing Runner: inception_4b/pool_proj + inception_4b/relu_pool_proj (CudnnConvolution) [TRT] Tactic: 0 Time: 0.241614 [TRT] Tactic: 1 Time: 0.214844 [TRT] Tactic: 2 Time: 0.541041 [TRT] Tactic: 4 skipped. Scratch requested: 76808192, available: 33554432 [TRT] Tactic: 5 Time: 2.2475 [TRT] Fastest Tactic: 1 Time: 0.214844 [TRT] --------------- Timing Runner: inception_4b/pool_proj + inception_4b/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4b/pool_proj + inception_4b/relu_pool_proj (CaskConvolution) [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.155052 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v0 Tactic: 1698681053543049347 [TRT] Tactic: 1698681053543049347 Time: 0.138256 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.199922 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.123125 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.198463 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v1 Tactic: 5326823351883942011 [TRT] Tactic: 5326823351883942011 Time: 0.191875 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v0 Tactic: 5500448035057547314 [TRT] Tactic: 5500448035057547314 Time: 0.149114 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.128411 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.145209 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.202552 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v1 Tactic: -6576203419454146580 [TRT] Tactic: -6576203419454146580 Time: 0.13888 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.151718 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.147058 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.160834 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.153516 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.196693 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v0 Tactic: -377491875521947884 [TRT] Tactic: -377491875521947884 Time: 0.197214 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v1 Tactic: -37215280111360163 [TRT] Tactic: -37215280111360163 Time: 0.120156 [TRT] Fastest Tactic: -37215280111360163 Time: 0.120156 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -37215280111360163 [TRT] *************** Autotuning format combination: Float(100352,1,7168,512) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: inception_4b/pool_proj + inception_4b/relu_pool_proj (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4b/pool_proj + inception_4b/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4b/pool_proj + inception_4b/relu_pool_proj (CaskConvolution) [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 3886731678879822788 [TRT] Tactic: 3886731678879822788 Time: 0.105417 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 6629944304117643200 [TRT] Tactic: 6629944304117643200 Time: 0.161458 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.161354 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.10625 [TRT] Fastest Tactic: 3886731678879822788 Time: 0.105417 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 3886731678879822788 [TRT] *************** Autotuning format combination: Half(100352,196,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/pool_proj + inception_4b/relu_pool_proj (CudnnConvolution) [TRT] Tactic: 0 Time: 0.272656 [TRT] Tactic: 1 Time: 0.255911 [TRT] Tactic: 2 Time: 0.534245 [TRT] Tactic: 4 skipped. Scratch requested: 76808192, available: 33554432 [TRT] Tactic: 5 Time: 2.09482 [TRT] Fastest Tactic: 1 Time: 0.255911 [TRT] --------------- Timing Runner: inception_4b/pool_proj + inception_4b/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4b/pool_proj + inception_4b/relu_pool_proj (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 1 [TRT] *************** Autotuning format combination: Half(50176,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/pool_proj + inception_4b/relu_pool_proj (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] *************** Autotuning format combination: Half(50176,196:2,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/pool_proj + inception_4b/relu_pool_proj (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.120677 [TRT] Tactic: 655359 Time: 0.206068 [TRT] Tactic: 786431 Time: 0.152891 [TRT] Tactic: 851967 Time: 0.187344 [TRT] Tactic: 1179647 Time: 0.095443 [TRT] Tactic: 1310719 Time: 0.263073 [TRT] Tactic: 1376255 Time: 0.121693 [TRT] Tactic: 1441791 Time: 0.128307 [TRT] Tactic: 1507327 Time: 0.238568 [TRT] Tactic: 1638399 Time: 0.156563 [TRT] Tactic: 1835007 Time: 0.145469 [TRT] Tactic: 1900543 Time: 0.220703 [TRT] Tactic: 2097151 Time: 0.152005 [TRT] Tactic: 2162687 Time: 0.124063 [TRT] Tactic: 2293759 Time: 0.123047 [TRT] Tactic: 2359295 Time: 0.128594 [TRT] Tactic: 2686975 Time: 0.217318 [TRT] Tactic: 3080191 Time: 0.161224 [TRT] Tactic: 3342335 Time: 0.187474 [TRT] Tactic: 3407871 Time: 0.107942 [TRT] Tactic: 3538943 Time: 0.095599 [TRT] Tactic: 3670015 Time: 0.220912 [TRT] Tactic: 3932159 Time: 0.177188 [TRT] Tactic: 3997695 Time: 0.138306 [TRT] Tactic: 4063231 Time: 0.165521 [TRT] Tactic: 4194303 Time: 0.114296 [TRT] Tactic: 4259839 Time: 0.150989 [TRT] Tactic: 4325375 Time: 0.14 [TRT] Tactic: 4521983 Time: 0.143516 [TRT] Tactic: 4587519 Time: 0.139505 [TRT] Tactic: 4653055 Time: 0.11552 [TRT] Tactic: 4915199 Time: 0.116354 [TRT] Tactic: 4980735 Time: 0.152448 [TRT] Tactic: 5177343 Time: 0.106146 [TRT] Tactic: 5242879 Time: 0.086875 [TRT] Tactic: 5373951 Time: 0.100572 [TRT] Tactic: 5439487 Time: 0.11315 [TRT] Tactic: 5570559 Time: 0.18039 [TRT] Tactic: 5636095 Time: 0.166822 [TRT] Tactic: 5701631 Time: 0.1175 [TRT] Tactic: 5767167 Time: 0.151198 [TRT] Tactic: 5832703 Time: 0.101797 [TRT] Tactic: 5898239 Time: 0.0964585 [TRT] Tactic: 6029311 Time: 0.116145 [TRT] Tactic: 6225919 Time: 0.0826305 [TRT] Tactic: 6291455 Time: 0.0953125 [TRT] Tactic: 6422527 Time: 0.152917 [TRT] Tactic: 6750207 Time: 0.113594 [TRT] Tactic: 6815743 Time: 0.0859895 [TRT] Tactic: 6946815 Time: 0.138698 [TRT] Tactic: 7012351 Time: 0.152422 [TRT] Tactic: 7077887 Time: 0.0901045 [TRT] Tactic: 7143423 Time: 0.162813 [TRT] Tactic: 7208959 Time: 0.106641 [TRT] Tactic: 7340031 Time: 0.0986195 [TRT] Tactic: 7405567 Time: 0.0955725 [TRT] Tactic: 7536639 Time: 0.115572 [TRT] Tactic: 7602175 Time: 0.128541 [TRT] Tactic: 7733247 Time: 0.095182 [TRT] Tactic: 7798783 Time: 0.152605 [TRT] Tactic: 8191999 Time: 0.167735 [TRT] Tactic: 8257535 Time: 0.116692 [TRT] Tactic: 8323071 Time: 0.110494 [TRT] Tactic: 8650751 Time: 0.140287 [TRT] Tactic: 8716287 Time: 0.085781 [TRT] Tactic: 9109503 Time: 0.150912 [TRT] Tactic: 9568255 Time: 0.116276 [TRT] Tactic: 9895935 Time: 0.113359 [TRT] Tactic: 10223615 Time: 0.220208 [TRT] Tactic: 10354687 Time: 0.151094 [TRT] Tactic: 10551295 Time: 0.103516 [TRT] Tactic: 10747903 Time: 0.0897655 [TRT] Tactic: 10944511 Time: 0.1525 [TRT] Fastest Tactic: 6225919 Time: 0.0826305 [TRT] --------------- Timing Runner: inception_4b/pool_proj + inception_4b/relu_pool_proj (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4b/pool_proj + inception_4b/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4b/pool_proj + inception_4b/relu_pool_proj (CaskConvolution) [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_interior_nn_v1 Tactic: 3066127711859985668 [TRT] Tactic: 3066127711859985668 Time: 0.0734895 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.0809895 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.0764845 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.0710935 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] Tactic: 8163473458334948789 Time: 0.0667445 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.105287 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.107526 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.068203 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_interior_nn_v1 Tactic: -1716393687483585322 [TRT] Tactic: -1716393687483585322 Time: 0.10276 [TRT] Fastest Tactic: 8163473458334948789 Time: 0.0667445 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 8163473458334948789 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(58016,196,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 2.5625 [TRT] Tactic: 0 Time: 0.0163805 [TRT] Fastest Tactic: 0 Time: 0.0163805 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(58016,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: inception_4b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0578385 [TRT] Tactic: 0 Time: 0.0567185 [TRT] Fastest Tactic: 0 Time: 0.0567185 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(58016,196,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 1.1168 [TRT] Tactic: 0 Time: 0.051614 [TRT] Fastest Tactic: 0 Time: 0.051614 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(58016,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.061537 [TRT] Tactic: 0 Time: 0.0617965 [TRT] Fastest Tactic: 1002 Time: 0.061537 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(58016,1,4144,296) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0704685 [TRT] Tactic: 0 Time: 0.052604 [TRT] Fastest Tactic: 0 Time: 0.052604 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(58016,1,4144,296) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: inception_4b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.086458 [TRT] Tactic: 0 Time: 0.0546355 [TRT] Fastest Tactic: 0 Time: 0.0546355 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(58016,1,4144,296) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.044167 [TRT] Tactic: 0 Time: 0.0517185 [TRT] Fastest Tactic: 1002 Time: 0.044167 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(58016,1,4144,296) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0534635 [TRT] Tactic: 0 Time: 0.0616665 [TRT] Fastest Tactic: 1002 Time: 0.0534635 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(58016,196,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 1.14339 [TRT] Tactic: 0 Time: 0.051667 [TRT] Fastest Tactic: 0 Time: 0.051667 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(58016,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: inception_4b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.040573 [TRT] Tactic: 0 Time: 0.0571355 [TRT] Fastest Tactic: 1002 Time: 0.040573 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(58016,196,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 1.13797 [TRT] Tactic: 0 Time: 0.0085155 [TRT] Fastest Tactic: 0 Time: 0.0085155 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(58016,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.045677 [TRT] Tactic: 0 Time: 0.0316405 [TRT] Fastest Tactic: 0 Time: 0.0316405 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(29008,196:2,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0574995 [TRT] Tactic: 0 Time: 0.059791 [TRT] Fastest Tactic: 1002 Time: 0.0574995 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(29008,196:2,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: inception_4b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.040625 [TRT] Tactic: 0 Time: 0.0648175 [TRT] Fastest Tactic: 1002 Time: 0.040625 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(29008,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.083099 [TRT] Tactic: 0 Time: 0.054869 [TRT] Fastest Tactic: 0 Time: 0.054869 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(29008,196:2,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0589845 [TRT] Tactic: 0 Time: 0.008412 [TRT] Fastest Tactic: 0 Time: 0.008412 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning format combination: Float(100352,196,14,1) -> Float(54880,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.894531 [TRT] Tactic: 655359 Time: 0.77151 [TRT] Tactic: 786431 Time: 0.962057 [TRT] Tactic: 851967 Time: 1.3138 [TRT] Tactic: 1179647 Time: 1.16362 [TRT] Tactic: 1310719 Time: 1.97417 [TRT] Tactic: 1376255 Time: 0.804739 [TRT] Tactic: 1441791 Time: 1.32 [TRT] Tactic: 1507327 Time: 1.31456 [TRT] Tactic: 1638399 Time: 1.41195 [TRT] Tactic: 1835007 Time: 0.983255 [TRT] Tactic: 1900543 Time: 1.24005 [TRT] Tactic: 2097151 Time: 1.09271 [TRT] Tactic: 2162687 Time: 0.832395 [TRT] Tactic: 2293759 Time: 0.885755 [TRT] Tactic: 2359295 Time: 1.04607 [TRT] Tactic: 2686975 Time: 0.966588 [TRT] Tactic: 3080191 Time: 0.93276 [TRT] Tactic: 3342335 Time: 1.21521 [TRT] Tactic: 3407871 Time: 0.916953 [TRT] Tactic: 3538943 Time: 0.993281 [TRT] Tactic: 3670015 Time: 0.765651 [TRT] Tactic: 3932159 Time: 1.44682 [TRT] Tactic: 3997695 Time: 0.960963 [TRT] Tactic: 4063231 Time: 1.15094 [TRT] Tactic: 4194303 Time: 0.941484 [TRT] Tactic: 4259839 Time: 1.12583 [TRT] Tactic: 4325375 Time: 1.18042 [TRT] Tactic: 4521983 Time: 1.22924 [TRT] Tactic: 4587519 Time: 1.0713 [TRT] Tactic: 4653055 Time: 1.12893 [TRT] Tactic: 4915199 Time: 0.907239 [TRT] Tactic: 4980735 Time: 1.17711 [TRT] Tactic: 5177343 Time: 1.37984 [TRT] Tactic: 5242879 Time: 0.848412 [TRT] Tactic: 5373951 Time: 1.29117 [TRT] Tactic: 5439487 Time: 0.961146 [TRT] Tactic: 5570559 Time: 0.905234 [TRT] Tactic: 5636095 Time: 1.14984 [TRT] Tactic: 5701631 Time: 1.05029 [TRT] Tactic: 5767167 Time: 1.46924 [TRT] Tactic: 5832703 Time: 0.911901 [TRT] Tactic: 5898239 Time: 0.827344 [TRT] Tactic: 6029311 Time: 0.825208 [TRT] Tactic: 6225919 Time: 0.898438 [TRT] Tactic: 6291455 Time: 1.16016 [TRT] Tactic: 6422527 Time: 0.994036 [TRT] Tactic: 6750207 Time: 0.908802 [TRT] Tactic: 6815743 Time: 0.911536 [TRT] Tactic: 6946815 Time: 1.16477 [TRT] Tactic: 7012351 Time: 1.0324 [TRT] Tactic: 7077887 Time: 0.956719 [TRT] Tactic: 7143423 Time: 1.29599 [TRT] Tactic: 7208959 Time: 1.02076 [TRT] Tactic: 7340031 Time: 0.842682 [TRT] Tactic: 7405567 Time: 0.957135 [TRT] Tactic: 7536639 Time: 1.00945 [TRT] Tactic: 7602175 Time: 1.08443 [TRT] Tactic: 7733247 Time: 0.848593 [TRT] Tactic: 7798783 Time: 0.921354 [TRT] Tactic: 8191999 Time: 1.34935 [TRT] Tactic: 8257535 Time: 0.882656 [TRT] Tactic: 8323071 Time: 0.792057 [TRT] Tactic: 8650751 Time: 1.23617 [TRT] Tactic: 8716287 Time: 0.942291 [TRT] Tactic: 9109503 Time: 1.10661 [TRT] Tactic: 9568255 Time: 0.846563 [TRT] Tactic: 9895935 Time: 0.875755 [TRT] Tactic: 10223615 Time: 0.904245 [TRT] Tactic: 10354687 Time: 1.10456 [TRT] Tactic: 10551295 Time: 0.796901 [TRT] Tactic: 10747903 Time: 0.803047 [TRT] Tactic: 10944511 Time: 1.09372 [TRT] Fastest Tactic: 3670015 Time: 0.765651 [TRT] --------------- Timing Runner: inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce (CudnnConvolution) [TRT] Tactic: 0 Time: 0.872292 [TRT] Tactic: 1 Time: 0.729818 [TRT] Tactic: 2 Time: 1.03102 [TRT] Tactic: 4 skipped. Scratch requested: 332054528, available: 33554432 [TRT] Tactic: 5 Time: 10.4304 [TRT] Fastest Tactic: 1 Time: 0.729818 [TRT] --------------- Timing Runner: inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce (CaskConvolution) [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.497083 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v0 Tactic: 1698681053543049347 [TRT] Tactic: 1698681053543049347 Time: 0.489558 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.541406 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.455703 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.534011 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v1 Tactic: 5326823351883942011 [TRT] Tactic: 5326823351883942011 Time: 0.515104 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v0 Tactic: 5500448035057547314 [TRT] Tactic: 5500448035057547314 Time: 0.508542 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.462291 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.556172 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.539818 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v1 Tactic: -6576203419454146580 [TRT] Tactic: -6576203419454146580 Time: 0.461224 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.583828 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.490287 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.532058 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.516771 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.531432 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v0 Tactic: -377491875521947884 [TRT] Tactic: -377491875521947884 Time: 0.527604 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v1 Tactic: -37215280111360163 [TRT] Tactic: -37215280111360163 Time: 0.441875 [TRT] Fastest Tactic: -37215280111360163 Time: 0.441875 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -37215280111360163 [TRT] *************** Autotuning format combination: Float(100352,1,7168,512) -> Float(54880,1,3920,280) *************** [TRT] --------------- Timing Runner: inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce (CaskConvolution) [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 3886731678879822788 [TRT] Tactic: 3886731678879822788 Time: 0.448646 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 6629944304117643200 [TRT] Tactic: 6629944304117643200 Time: 0.676224 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.679193 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.456537 [TRT] Fastest Tactic: 3886731678879822788 Time: 0.448646 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 3886731678879822788 [TRT] *************** Autotuning format combination: Half(100352,196,14,1) -> Half(54880,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce (CudnnConvolution) [TRT] Tactic: 0 Time: 0.949089 [TRT] Tactic: 1 Time: 0.765703 [TRT] Tactic: 2 Time: 1.01273 [TRT] Tactic: 4 skipped. Scratch requested: 332054528, available: 33554432 [TRT] Tactic: 5 Time: 10.8469 [TRT] Fastest Tactic: 1 Time: 0.765703 [TRT] --------------- Timing Runner: inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 1 [TRT] *************** Autotuning format combination: Half(50176,196:2,14,1) -> Half(54880,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] *************** Autotuning format combination: Half(50176,196:2,14,1) -> Half(27440,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.403464 [TRT] Tactic: 655359 Time: 0.494636 [TRT] Tactic: 786431 Time: 0.576276 [TRT] Tactic: 851967 Time: 0.671198 [TRT] Tactic: 1179647 Time: 0.46112 [TRT] Tactic: 1310719 Time: 0.928698 [TRT] Tactic: 1376255 Time: 0.373515 [TRT] Tactic: 1441791 Time: 0.621771 [TRT] Tactic: 1507327 Time: 0.676979 [TRT] Tactic: 1638399 Time: 0.696146 [TRT] Tactic: 1835007 Time: 0.582813 [TRT] Tactic: 1900543 Time: 0.622162 [TRT] Tactic: 2097151 Time: 0.696667 [TRT] Tactic: 2162687 Time: 0.391537 [TRT] Tactic: 2293759 Time: 0.410755 [TRT] Tactic: 2359295 Time: 0.503281 [TRT] Tactic: 2686975 Time: 0.705365 [TRT] Tactic: 3080191 Time: 0.512136 [TRT] Tactic: 3342335 Time: 0.621198 [TRT] Tactic: 3407871 Time: 0.415807 [TRT] Tactic: 3538943 Time: 0.453307 [TRT] Tactic: 3670015 Time: 0.470547 [TRT] Tactic: 3932159 Time: 0.60349 [TRT] Tactic: 3997695 Time: 0.571849 [TRT] Tactic: 4063231 Time: 0.570078 [TRT] Tactic: 4194303 Time: 0.493047 [TRT] Tactic: 4259839 Time: 0.690469 [TRT] Tactic: 4325375 Time: 0.5675 [TRT] Tactic: 4521983 Time: 0.541147 [TRT] Tactic: 4587519 Time: 0.58638 [TRT] Tactic: 4653055 Time: 0.551041 [TRT] Tactic: 4915199 Time: 0.49112 [TRT] Tactic: 4980735 Time: 0.562422 [TRT] Tactic: 5177343 Time: 0.566041 [TRT] Tactic: 5242879 Time: 0.378099 [TRT] Tactic: 5373951 Time: 0.551145 [TRT] Tactic: 5439487 Time: 0.51 [TRT] Tactic: 5570559 Time: 0.542292 [TRT] Tactic: 5636095 Time: 0.570729 [TRT] Tactic: 5701631 Time: 0.445886 [TRT] Tactic: 5767167 Time: 0.630234 [TRT] Tactic: 5832703 Time: 0.409375 [TRT] Tactic: 5898239 Time: 0.465912 [TRT] Tactic: 6029311 Time: 0.371745 [TRT] Tactic: 6225919 Time: 0.39836 [TRT] Tactic: 6291455 Time: 0.463125 [TRT] Tactic: 6422527 Time: 0.461172 [TRT] Tactic: 6750207 Time: 0.500808 [TRT] Tactic: 6815743 Time: 0.410912 [TRT] Tactic: 6946815 Time: 0.575443 [TRT] Tactic: 7012351 Time: 0.697708 [TRT] Tactic: 7077887 Time: 0.432031 [TRT] Tactic: 7143423 Time: 0.65552 [TRT] Tactic: 7208959 Time: 0.476172 [TRT] Tactic: 7340031 Time: 0.503281 [TRT] Tactic: 7405567 Time: 0.47513 [TRT] Tactic: 7536639 Time: 0.493724 [TRT] Tactic: 7602175 Time: 0.523046 [TRT] Tactic: 7733247 Time: 0.469505 [TRT] Tactic: 7798783 Time: 0.578256 [TRT] Tactic: 8191999 Time: 0.685182 [TRT] Tactic: 8257535 Time: 0.50651 [TRT] Tactic: 8323071 Time: 0.463151 [TRT] Tactic: 8650751 Time: 0.611745 [TRT] Tactic: 8716287 Time: 0.455104 [TRT] Tactic: 9109503 Time: 0.706199 [TRT] Tactic: 9568255 Time: 0.494011 [TRT] Tactic: 9895935 Time: 0.493047 [TRT] Tactic: 10223615 Time: 0.704948 [TRT] Tactic: 10354687 Time: 0.689844 [TRT] Tactic: 10551295 Time: 0.404401 [TRT] Tactic: 10747903 Time: 0.409792 [TRT] Tactic: 10944511 Time: 0.561927 [TRT] Fastest Tactic: 6029311 Time: 0.371745 [TRT] --------------- Timing Runner: inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce (CaskConvolution) [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_interior_nn_v1 Tactic: 3066127711859985668 [TRT] Tactic: 3066127711859985668 Time: 0.241536 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.277292 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.245209 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.241666 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] Tactic: 8163473458334948789 Time: 0.233723 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.277578 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.279219 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.234583 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_interior_nn_v1 Tactic: -1716393687483585322 [TRT] Tactic: -1716393687483585322 Time: 0.272214 [TRT] Fastest Tactic: 8163473458334948789 Time: 0.233723 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 8163473458334948789 [TRT] *************** Autotuning Reformat:Float(54880,196,14,1) -> Float(54880,1,3920,280) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0871355 [TRT] Tactic: 0 Time: 0.0882555 [TRT] Fastest Tactic: 1002 Time: 0.0871355 [TRT] *************** Autotuning Reformat:Float(54880,196,14,1) -> Half(54880,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.47911 [TRT] Tactic: 0 Time: 0.0625 [TRT] Fastest Tactic: 0 Time: 0.0625 [TRT] *************** Autotuning Reformat:Float(54880,196,14,1) -> Half(27440,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.092865 [TRT] Tactic: 0 Time: 0.049167 [TRT] Fastest Tactic: 0 Time: 0.049167 [TRT] *************** Autotuning Reformat:Float(54880,1,3920,280) -> Float(54880,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.108516 [TRT] Tactic: 0 Time: 0.080755 [TRT] Fastest Tactic: 0 Time: 0.080755 [TRT] *************** Autotuning Reformat:Float(54880,1,3920,280) -> Half(54880,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.068594 [TRT] Tactic: 0 Time: 0.0810155 [TRT] Fastest Tactic: 1002 Time: 0.068594 [TRT] *************** Autotuning Reformat:Float(54880,1,3920,280) -> Half(27440,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0815625 [TRT] Tactic: 0 Time: 0.0959635 [TRT] Fastest Tactic: 1002 Time: 0.0815625 [TRT] *************** Autotuning Reformat:Half(54880,196,14,1) -> Float(54880,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.52518 [TRT] Tactic: 0 Time: 0.0521095 [TRT] Fastest Tactic: 0 Time: 0.0521095 [TRT] *************** Autotuning Reformat:Half(54880,196,14,1) -> Float(54880,1,3920,280) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0634115 [TRT] Tactic: 0 Time: 0.0874735 [TRT] Fastest Tactic: 1002 Time: 0.0634115 [TRT] *************** Autotuning Reformat:Half(54880,196,14,1) -> Half(27440,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0672395 [TRT] Tactic: 0 Time: 0.0490105 [TRT] Fastest Tactic: 0 Time: 0.0490105 [TRT] *************** Autotuning Reformat:Half(27440,196:2,14,1) -> Float(54880,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0882285 [TRT] Tactic: 0 Time: 0.043333 [TRT] Fastest Tactic: 0 Time: 0.043333 [TRT] *************** Autotuning Reformat:Half(27440,196:2,14,1) -> Float(54880,1,3920,280) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0644275 [TRT] Tactic: 0 Time: 0.101328 [TRT] Fastest Tactic: 1002 Time: 0.0644275 [TRT] *************** Autotuning Reformat:Half(27440,196:2,14,1) -> Half(54880,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.125313 [TRT] Tactic: 0 Time: 0.042031 [TRT] Fastest Tactic: 0 Time: 0.042031 [TRT] *************** Autotuning Reformat:Float(54880,196,14,1) -> Float(54880,1,3920,280) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.035729 [TRT] Tactic: 0 Time: 0.040938 [TRT] Fastest Tactic: 1002 Time: 0.035729 [TRT] *************** Autotuning Reformat:Float(54880,196,14,1) -> Half(54880,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.82 [TRT] Tactic: 0 Time: 0.038021 [TRT] Fastest Tactic: 0 Time: 0.038021 [TRT] *************** Autotuning Reformat:Float(54880,196,14,1) -> Half(27440,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.046276 [TRT] Tactic: 0 Time: 0.0451305 [TRT] Fastest Tactic: 0 Time: 0.0451305 [TRT] *************** Autotuning Reformat:Float(54880,1,3920,280) -> Float(54880,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.047344 [TRT] Tactic: 0 Time: 0.0382815 [TRT] Fastest Tactic: 0 Time: 0.0382815 [TRT] *************** Autotuning Reformat:Float(54880,1,3920,280) -> Half(54880,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.033568 [TRT] Tactic: 0 Time: 0.038099 [TRT] Fastest Tactic: 1002 Time: 0.033568 [TRT] *************** Autotuning Reformat:Float(54880,1,3920,280) -> Half(27440,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0403905 [TRT] Tactic: 0 Time: 0.0451305 [TRT] Fastest Tactic: 1002 Time: 0.0403905 [TRT] *************** Autotuning Reformat:Half(54880,196,14,1) -> Float(54880,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.839609 [TRT] Tactic: 0 Time: 0.026407 [TRT] Fastest Tactic: 0 Time: 0.026407 [TRT] *************** Autotuning Reformat:Half(54880,196,14,1) -> Float(54880,1,3920,280) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.031146 [TRT] Tactic: 0 Time: 0.042995 [TRT] Fastest Tactic: 1002 Time: 0.031146 [TRT] *************** Autotuning Reformat:Half(54880,196,14,1) -> Half(27440,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0340365 [TRT] Tactic: 0 Time: 0.024036 [TRT] Fastest Tactic: 0 Time: 0.024036 [TRT] *************** Autotuning Reformat:Half(27440,196:2,14,1) -> Float(54880,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.044297 [TRT] Tactic: 0 Time: 0.021771 [TRT] Fastest Tactic: 0 Time: 0.021771 [TRT] *************** Autotuning Reformat:Half(27440,196:2,14,1) -> Float(54880,1,3920,280) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.030781 [TRT] Tactic: 0 Time: 0.047474 [TRT] Fastest Tactic: 1002 Time: 0.030781 [TRT] *************** Autotuning Reformat:Half(27440,196:2,14,1) -> Half(54880,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.058307 [TRT] Tactic: 0 Time: 0.040365 [TRT] Fastest Tactic: 0 Time: 0.040365 [TRT] *************** Autotuning format combination: Float(54880,196,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4c/3x3 + inception_4c/relu_3x3 (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4c/3x3 + inception_4c/relu_3x3 (FusedConvActConvolution) [TRT] Tactic: 524287 Time: 0.822474 [TRT] Tactic: 720895 Time: 0.896589 [TRT] Tactic: 983039 Time: 0.762266 [TRT] Tactic: 1048575 Time: 0.845052 [TRT] Tactic: 1703935 Time: 0.825312 [TRT] Tactic: 1769471 Time: 0.795443 [TRT] Tactic: 1966079 Time: 0.840052 [TRT] Tactic: 2031615 Time: 0.821276 [TRT] Tactic: 2228223 Time: 1.06687 [TRT] Tactic: 2424831 Time: 1.16682 [TRT] Tactic: 2621439 Time: 0.996172 [TRT] Tactic: 2752511 Time: 0.739999 [TRT] Tactic: 2818047 Time: 1.04208 [TRT] Tactic: 2883583 Time: 0.811381 [TRT] Tactic: 3014655 Time: 0.776693 [TRT] Tactic: 3145727 Time: 0.728463 [TRT] Tactic: 3473407 Time: 0.89138 [TRT] Tactic: 3604479 Time: 0.766511 [TRT] Tactic: 3735551 Time: 1.03008 [TRT] Tactic: 4390911 Time: 0.751354 [TRT] Tactic: 5046271 Time: 0.805624 [TRT] Tactic: 5963775 Time: 0.770729 [TRT] Tactic: 6160383 Time: 0.808333 [TRT] Tactic: 6488063 Time: 0.712656 [TRT] Tactic: 6881279 Time: 0.767787 [TRT] Tactic: 7274495 Time: 1.00586 [TRT] Tactic: 7864319 Time: 1.04451 [TRT] Tactic: 7995391 Time: 0.86862 [TRT] Tactic: 8585215 Time: 0.741875 [TRT] Tactic: 8847359 Time: 0.77724 [TRT] Tactic: 8978431 Time: 0.775911 [TRT] Tactic: 9043967 Time: 0.703776 [TRT] Tactic: 9175039 Time: 0.766484 [TRT] Tactic: 9502719 Time: 0.764739 [TRT] Tactic: 9830399 Time: 0.783255 [TRT] Tactic: 9961471 Time: 0.865469 [TRT] Tactic: 10027007 Time: 0.728619 [TRT] Tactic: 10092543 Time: 0.739115 [TRT] Tactic: 10289151 Time: 0.848907 [TRT] Tactic: 10485759 Time: 0.725183 [TRT] Tactic: 10682367 Time: 1.00013 [TRT] Tactic: 10813439 Time: 0.873151 [TRT] Fastest Tactic: 9043967 Time: 0.703776 [TRT] --------------- Timing Runner: inception_4c/3x3 + inception_4c/relu_3x3 (CudnnConvolution) [TRT] Tactic: 0 Time: 1.52661 [TRT] Tactic: 1 Time: 0.964193 [TRT] Tactic: 2 Time: 1.37899 [TRT] Tactic: 4 skipped. Scratch requested: 76972032, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 144277504, available: 33554432 [TRT] Tactic: 6 Time: 0.744974 [TRT] Fastest Tactic: 6 Time: 0.744974 [TRT] --------------- Timing Runner: inception_4c/3x3 + inception_4c/relu_3x3 (CaskConvolution) [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 1.03784 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v0 Tactic: 1754984623894446479 [TRT] Tactic: 1754984623894446479 Time: 1.30101 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v0 Tactic: 3611739942397549984 [TRT] Tactic: 3611739942397549984 Time: 0.792812 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148n_nt_v1 Tactic: 3827454225649558724 [TRT] Tactic: 3827454225649558724 Time: 0.656796 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v1 Tactic: 4337000649858996379 [TRT] Tactic: 4337000649858996379 Time: 0.814479 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.794011 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.765547 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.773229 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 5921334924264294896 [TRT] Tactic: 5921334924264294896 Time: 0.496459 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.800182 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 1.11255 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148t_nt_v1 Tactic: 7852627285308570038 [TRT] Tactic: 7852627285308570038 Time: 0.636302 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v1 Tactic: -9137461792520977713 [TRT] Tactic: -9137461792520977713 Time: 0.79888 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148n_nt_v0 Tactic: -8776506421218919509 [TRT] Tactic: -8776506421218919509 Time: 0.622057 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.79625 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v0 Tactic: -8133971918129952780 [TRT] Tactic: -8133971918129952780 Time: 0.864349 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v1 Tactic: -6092040395344634144 [TRT] Tactic: -6092040395344634144 Time: 1.08146 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 1.29768 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.900312 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148t_nt_v0 Tactic: -2318106587342035239 [TRT] Tactic: -2318106587342035239 Time: 0.611874 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_mobile_relu_tile148t_nt_v0 Tactic: -1343271414618805657 [TRT] Tactic: -1343271414618805657 Time: 0.454376 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.859895 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.828359 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.762239 [TRT] Fastest Tactic: -1343271414618805657 Time: 0.454376 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -1343271414618805657 [TRT] *************** Autotuning format combination: Float(54880,1,3920,280) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: inception_4c/3x3 + inception_4c/relu_3x3 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4c/3x3 + inception_4c/relu_3x3 (CaskConvolution) [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.971953 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.753541 [TRT] Fastest Tactic: -7394439838318485025 Time: 0.753541 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(54880,196,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4c/3x3 + inception_4c/relu_3x3 (CudnnConvolution) [TRT] Tactic: 0 Time: 1.83177 [TRT] Tactic: 1 Time: 1.86997 [TRT] Tactic: 2 Time: 1.3494 [TRT] Tactic: 4 skipped. Scratch requested: 76972032, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 144277504, available: 33554432 [TRT] Tactic: 6 Time: 2.12891 [TRT] Fastest Tactic: 2 Time: 1.3494 [TRT] --------------- Timing Runner: inception_4c/3x3 + inception_4c/relu_3x3 (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 2 [TRT] *************** Autotuning format combination: Half(27440,196:2,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4c/3x3 + inception_4c/relu_3x3 (FusedConvActConvolution) [TRT] Tactic: 524287 Time: 0.436068 [TRT] Tactic: 720895 Time: 0.540573 [TRT] Tactic: 983039 Time: 0.428073 [TRT] Tactic: 1048575 Time: 0.468958 [TRT] Tactic: 1703935 Time: 0.475755 [TRT] Tactic: 1769471 Time: 2.70417 [TRT] Tactic: 1966079 Time: 0.494999 [TRT] Tactic: 2031615 Time: 0.426901 [TRT] Tactic: 2228223 Time: 0.591718 [TRT] Tactic: 2424831 Time: 0.876276 [TRT] Tactic: 2621439 Time: 0.551224 [TRT] Tactic: 2752511 Time: 0.432162 [TRT] Tactic: 2818047 Time: 0.572083 [TRT] Tactic: 2883583 Time: 0.520989 [TRT] Tactic: 3014655 Time: 0.447084 [TRT] Tactic: 3145727 Time: 0.428489 [TRT] Tactic: 3473407 Time: 0.522812 [TRT] Tactic: 3604479 Time: 0.441927 [TRT] Tactic: 3735551 Time: 0.492709 [TRT] Tactic: 4390911 Time: 0.426145 [TRT] Tactic: 5046271 Time: 0.438672 [TRT] Tactic: 5963775 Time: 0.421276 [TRT] Tactic: 6160383 Time: 0.464714 [TRT] Tactic: 6488063 Time: 0.403176 [TRT] Tactic: 6881279 Time: 0.398021 [TRT] Tactic: 7274495 Time: 0.567266 [TRT] Tactic: 7864319 Time: 0.587448 [TRT] Tactic: 7995391 Time: 0.52573 [TRT] Tactic: 8585215 Time: 0.421121 [TRT] Tactic: 8847359 Time: 0.438334 [TRT] Tactic: 8978431 Time: 0.442708 [TRT] Tactic: 9043967 Time: 0.394713 [TRT] Tactic: 9175039 Time: 0.440234 [TRT] Tactic: 9502719 Time: 0.425234 [TRT] Tactic: 9830399 Time: 0.407995 [TRT] Tactic: 9961471 Time: 0.487916 [TRT] Tactic: 10027007 Time: 0.403203 [TRT] Tactic: 10092543 Time: 0.430052 [TRT] Tactic: 10289151 Time: 0.500208 [TRT] Tactic: 10485759 Time: 0.399062 [TRT] Tactic: 10682367 Time: 0.53763 [TRT] Tactic: 10813439 Time: 0.532735 [TRT] Fastest Tactic: 9043967 Time: 0.394713 [TRT] --------------- Timing Runner: inception_4c/3x3 + inception_4c/relu_3x3 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4c/3x3 + inception_4c/relu_3x3 (CaskConvolution) [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.531979 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_large_nn_v1 Tactic: 3650389455493082349 [TRT] Tactic: 3650389455493082349 Time: 0.564246 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_winograd_fp16x2_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 4772821744921268633 [TRT] Tactic: 4772821744921268633 Time: 0.266719 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.456172 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.415522 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_large_nn_v1 Tactic: -6490690591794140522 [TRT] Tactic: -6490690591794140522 Time: 0.420885 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_large_nn_v1 Tactic: -4686027666808657977 [TRT] Tactic: -4686027666808657977 Time: 0.406849 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.387448 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.404531 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.403646 [TRT] Fastest Tactic: 4772821744921268633 Time: 0.266719 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 4772821744921268633 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.072318 [TRT] Tactic: 0 Time: 0.0825005 [TRT] Fastest Tactic: 1002 Time: 0.072318 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.63292 [TRT] Tactic: 0 Time: 0.0741665 [TRT] Fastest Tactic: 0 Time: 0.0741665 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.085703 [TRT] Tactic: 0 Time: 0.0453125 [TRT] Fastest Tactic: 0 Time: 0.0453125 [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.093385 [TRT] Tactic: 0 Time: 0.0764845 [TRT] Fastest Tactic: 0 Time: 0.0764845 [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0626045 [TRT] Tactic: 0 Time: 0.074817 [TRT] Fastest Tactic: 1002 Time: 0.0626045 [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.075026 [TRT] Tactic: 0 Time: 0.087969 [TRT] Fastest Tactic: 1002 Time: 0.075026 [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.67047 [TRT] Tactic: 0 Time: 0.075 [TRT] Fastest Tactic: 0 Time: 0.075 [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0565885 [TRT] Tactic: 0 Time: 0.0798695 [TRT] Fastest Tactic: 1002 Time: 0.0565885 [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.061536 [TRT] Tactic: 0 Time: 0.045156 [TRT] Fastest Tactic: 0 Time: 0.045156 [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.082396 [TRT] Tactic: 0 Time: 0.0871355 [TRT] Fastest Tactic: 1002 Time: 0.082396 [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0573435 [TRT] Tactic: 0 Time: 0.0920835 [TRT] Fastest Tactic: 1002 Time: 0.0573435 [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.109635 [TRT] Tactic: 0 Time: 0.0776305 [TRT] Fastest Tactic: 0 Time: 0.0776305 [TRT] *************** Autotuning Reformat:Float(54880,196,14,1) -> Float(54880,1,3920,280) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0241925 [TRT] Tactic: 0 Time: 0.0102345 [TRT] Fastest Tactic: 0 Time: 0.0102345 [TRT] *************** Autotuning Reformat:Float(54880,196,14,1) -> Half(54880,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.158489 [TRT] Tactic: 0 Time: 0.01026 [TRT] Fastest Tactic: 0 Time: 0.01026 [TRT] *************** Autotuning Reformat:Float(54880,196,14,1) -> Half(27440,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.013776 [TRT] Tactic: 0 Time: 0.012526 [TRT] Fastest Tactic: 0 Time: 0.012526 [TRT] *************** Autotuning Reformat:Float(54880,1,3920,280) -> Float(54880,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0254425 [TRT] Tactic: 0 Time: 0.010338 [TRT] Fastest Tactic: 0 Time: 0.010338 [TRT] *************** Autotuning Reformat:Float(54880,1,3920,280) -> Half(54880,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.012656 [TRT] Tactic: 0 Time: 0.010182 [TRT] Fastest Tactic: 0 Time: 0.010182 [TRT] *************** Autotuning Reformat:Float(54880,1,3920,280) -> Half(27440,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.012813 [TRT] Tactic: 0 Time: 0.0126045 [TRT] Fastest Tactic: 0 Time: 0.0126045 [TRT] *************** Autotuning Reformat:Half(54880,196,14,1) -> Float(54880,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.163203 [TRT] Tactic: 0 Time: 0.0078125 [TRT] Fastest Tactic: 0 Time: 0.0078125 [TRT] *************** Autotuning Reformat:Half(54880,196,14,1) -> Float(54880,1,3920,280) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.012526 [TRT] Tactic: 0 Time: 0.010312 [TRT] Fastest Tactic: 0 Time: 0.010312 [TRT] *************** Autotuning Reformat:Half(54880,196,14,1) -> Half(27440,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0127085 [TRT] Tactic: 0 Time: 0.0079685 [TRT] Fastest Tactic: 0 Time: 0.0079685 [TRT] *************** Autotuning Reformat:Half(27440,196:2,14,1) -> Float(54880,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.015 [TRT] Tactic: 0 Time: 0.007839 [TRT] Fastest Tactic: 0 Time: 0.007839 [TRT] *************** Autotuning Reformat:Half(27440,196:2,14,1) -> Float(54880,1,3920,280) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0116405 [TRT] Tactic: 0 Time: 0.0126045 [TRT] Fastest Tactic: 1002 Time: 0.0116405 [TRT] *************** Autotuning Reformat:Half(27440,196:2,14,1) -> Half(54880,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.024141 [TRT] Tactic: 0 Time: 0.01026 [TRT] Fastest Tactic: 0 Time: 0.01026 [TRT] *************** Autotuning format combination: Float(54880,196,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4c/5x5 + inception_4c/relu_5x5 (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4c/5x5 + inception_4c/relu_5x5 (FusedConvActConvolution) [TRT] Tactic: 393215 Time: 0.137448 [TRT] Tactic: 917503 Time: 0.125938 [TRT] Tactic: 1114111 Time: 0.12914 [TRT] Tactic: 1245183 Time: 0.13401 [TRT] Tactic: 1572863 Time: 0.134505 [TRT] Tactic: 2490367 Time: 0.150937 [TRT] Tactic: 2555903 Time: 0.141171 [TRT] Tactic: 2949119 Time: 0.117005 [TRT] Tactic: 3211263 Time: 0.221719 [TRT] Tactic: 3801087 Time: 0.12789 [TRT] Tactic: 3866623 Time: 0.129844 [TRT] Tactic: 4128767 Time: 0.114505 [TRT] Tactic: 4456447 Time: 0.116822 [TRT] Tactic: 4718591 Time: 0.116536 [TRT] Tactic: 4784127 Time: 0.2975 [TRT] Tactic: 4849663 Time: 0.120833 [TRT] Tactic: 5111807 Time: 0.120834 [TRT] Tactic: 5308415 Time: 0.140182 [TRT] Tactic: 5505023 Time: 0.200182 [TRT] Tactic: 6094847 Time: 0.132865 [TRT] Tactic: 6356991 Time: 0.14112 [TRT] Tactic: 6553599 Time: 0.138437 [TRT] Tactic: 6619135 Time: 0.150937 [TRT] Tactic: 6684671 Time: 0.21125 [TRT] Tactic: 7471103 Time: 0.131901 [TRT] Tactic: 7667711 Time: 0.117917 [TRT] Tactic: 7929855 Time: 0.128074 [TRT] Tactic: 8060927 Time: 0.142891 [TRT] Tactic: 8126463 Time: 0.125026 [TRT] Tactic: 8388607 Time: 0.137891 [TRT] Tactic: 8519679 Time: 0.155859 [TRT] Tactic: 8781823 Time: 0.144818 [TRT] Tactic: 8912895 Time: 0.123333 [TRT] Tactic: 9240575 Time: 0.114531 [TRT] Tactic: 9306111 Time: 0.193776 [TRT] Tactic: 9371647 Time: 0.125495 [TRT] Tactic: 9437183 Time: 0.115078 [TRT] Tactic: 9633791 Time: 0.121146 [TRT] Tactic: 9699327 Time: 0.1375 [TRT] Tactic: 9764863 Time: 0.143385 [TRT] Tactic: 10158079 Time: 0.142239 [TRT] Tactic: 10420223 Time: 0.128125 [TRT] Tactic: 10616831 Time: 0.128619 [TRT] Tactic: 10878975 Time: 0.153151 [TRT] Fastest Tactic: 4128767 Time: 0.114505 [TRT] --------------- Timing Runner: inception_4c/5x5 + inception_4c/relu_5x5 (CudnnConvolution) [TRT] Tactic: 0 Time: 0.278829 [TRT] Tactic: 1 Time: 0.262968 [TRT] Tactic: 2 Time: 0.439088 [TRT] Tactic: 4 Time: 2.80622 [TRT] Tactic: 5 Time: 2.59841 [TRT] Fastest Tactic: 1 Time: 0.262968 [TRT] --------------- Timing Runner: inception_4c/5x5 + inception_4c/relu_5x5 (CaskConvolution) [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.190339 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v0 Tactic: 1754984623894446479 [TRT] Tactic: 1754984623894446479 Time: 0.18474 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v0 Tactic: 3611739942397549984 [TRT] Tactic: 3611739942397549984 Time: 0.223438 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v1 Tactic: 4337000649858996379 [TRT] Tactic: 4337000649858996379 Time: 0.152422 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.224817 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.127395 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.20789 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.147526 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.141901 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v1 Tactic: -9137461792520977713 [TRT] Tactic: -9137461792520977713 Time: 0.227291 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.227161 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v0 Tactic: -8133971918129952780 [TRT] Tactic: -8133971918129952780 Time: 0.155026 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v1 Tactic: -6092040395344634144 [TRT] Tactic: -6092040395344634144 Time: 0.193151 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.184271 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.148125 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.155365 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.134453 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.208515 [TRT] Fastest Tactic: 5137655947464784826 Time: 0.127395 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 4128767 [TRT] *************** Autotuning format combination: Float(54880,1,3920,280) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: inception_4c/5x5 + inception_4c/relu_5x5 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4c/5x5 + inception_4c/relu_5x5 (CaskConvolution) [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.186224 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.140781 [TRT] Fastest Tactic: -7394439838318485025 Time: 0.140781 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(54880,196,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4c/5x5 + inception_4c/relu_5x5 (CudnnConvolution) [TRT] Tactic: 0 Time: 0.283906 [TRT] Tactic: 1 Time: 0.29875 [TRT] Tactic: 2 Time: 0.434427 [TRT] Tactic: 4 Time: 2.80323 [TRT] Tactic: 5 Time: 2.6306 [TRT] Fastest Tactic: 0 Time: 0.283906 [TRT] --------------- Timing Runner: inception_4c/5x5 + inception_4c/relu_5x5 (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 0 [TRT] *************** Autotuning format combination: Half(27440,196:2,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4c/5x5 + inception_4c/relu_5x5 (FusedConvActConvolution) [TRT] Tactic: 393215 Time: 0.078698 [TRT] Tactic: 917503 Time: 0.076146 [TRT] Tactic: 1114111 Time: 0.0687765 [TRT] Tactic: 1245183 Time: 0.073203 [TRT] Tactic: 1572863 Time: 0.0815105 [TRT] Tactic: 3211263 Time: 0.13427 [TRT] Tactic: 3801087 Time: 0.0733595 [TRT] Tactic: 3866623 Time: 0.075208 [TRT] Tactic: 4128767 Time: 0.0661985 [TRT] Tactic: 4456447 Time: 0.073047 [TRT] Tactic: 4718591 Time: 0.065521 [TRT] Tactic: 4784127 Time: 0.153333 [TRT] Tactic: 4849663 Time: 0.068151 [TRT] Tactic: 5111807 Time: 0.067656 [TRT] Tactic: 5308415 Time: 0.081328 [TRT] Tactic: 6094847 Time: 0.075729 [TRT] Tactic: 6553599 Time: 0.078802 [TRT] Tactic: 6619135 Time: 0.078203 [TRT] Tactic: 7471103 Time: 0.0780725 [TRT] Tactic: 7667711 Time: 0.0654425 [TRT] Tactic: 7929855 Time: 0.067031 [TRT] Tactic: 8781823 Time: 0.075703 [TRT] Tactic: 9240575 Time: 0.061302 [TRT] Tactic: 9306111 Time: 0.100157 [TRT] Tactic: 9371647 Time: 0.0685935 [TRT] Tactic: 9633791 Time: 0.066745 [TRT] Tactic: 9699327 Time: 0.08013 [TRT] Tactic: 9764863 Time: 0.0876825 [TRT] Tactic: 10158079 Time: 0.0812765 [TRT] Tactic: 10616831 Time: 0.0731515 [TRT] Tactic: 10878975 Time: 0.0893225 [TRT] Fastest Tactic: 9240575 Time: 0.061302 [TRT] --------------- Timing Runner: inception_4c/5x5 + inception_4c/relu_5x5 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4c/5x5 + inception_4c/relu_5x5 (CaskConvolution) [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.100885 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_large_nn_v1 Tactic: 3650389455493082349 [TRT] Tactic: 3650389455493082349 Time: 0.106798 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.0809375 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.0825 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_large_nn_v1 Tactic: -6490690591794140522 [TRT] Tactic: -6490690591794140522 Time: 0.082604 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_large_nn_v1 Tactic: -4686027666808657977 [TRT] Tactic: -4686027666808657977 Time: 0.120911 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.1125 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.120287 [TRT] inception_4c/5x5 + inception_4c/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.0686195 [TRT] Fastest Tactic: -2409163523992614473 Time: 0.0686195 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 9240575 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning format combination: Float(100352,196,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning format combination: Half(100352,196,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning format combination: Half(50176,196:2,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning format combination: Float(100352,196,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning format combination: Float(100352,1,7168,512) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning format combination: Half(100352,196,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning format combination: Half(50176,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4c/pool_proj + inception_4c/relu_pool_proj (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] *************** Autotuning format combination: Half(50176,196:2,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(54880,196,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4c/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 1.8788 [TRT] Tactic: 0 Time: 0.0102085 [TRT] Fastest Tactic: 0 Time: 0.0102085 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(54880,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: inception_4c/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0360415 [TRT] Tactic: 0 Time: 0.0422135 [TRT] Fastest Tactic: 1002 Time: 0.0360415 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(54880,196,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4c/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.820001 [TRT] Tactic: 0 Time: 0.038047 [TRT] Fastest Tactic: 0 Time: 0.038047 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(54880,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4c/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.045911 [TRT] Tactic: 0 Time: 0.045078 [TRT] Fastest Tactic: 0 Time: 0.045078 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(54880,1,3920,280) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4c/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0473955 [TRT] Tactic: 0 Time: 0.038906 [TRT] Fastest Tactic: 0 Time: 0.038906 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(54880,1,3920,280) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: inception_4c/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0649225 [TRT] Tactic: 0 Time: 0.041198 [TRT] Fastest Tactic: 0 Time: 0.041198 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(54880,1,3920,280) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4c/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.033489 [TRT] Tactic: 0 Time: 0.0385675 [TRT] Fastest Tactic: 1002 Time: 0.033489 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(54880,1,3920,280) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4c/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0403385 [TRT] Tactic: 0 Time: 0.0454945 [TRT] Fastest Tactic: 1002 Time: 0.0403385 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(54880,196,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4c/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.839453 [TRT] Tactic: 0 Time: 0.0380725 [TRT] Fastest Tactic: 0 Time: 0.0380725 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(54880,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: inception_4c/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.031094 [TRT] Tactic: 0 Time: 0.0423965 [TRT] Fastest Tactic: 1002 Time: 0.031094 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(54880,196,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4c/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.835 [TRT] Tactic: 0 Time: 0.0057815 [TRT] Fastest Tactic: 0 Time: 0.0057815 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(54880,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4c/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0340365 [TRT] Tactic: 0 Time: 0.024062 [TRT] Fastest Tactic: 0 Time: 0.024062 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(27440,196:2,14,1) -> Float(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4c/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.045052 [TRT] Tactic: 0 Time: 0.044948 [TRT] Fastest Tactic: 0 Time: 0.044948 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(27440,196:2,14,1) -> Float(100352,1,7168,512) *************** [TRT] --------------- Timing Runner: inception_4c/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0311195 [TRT] Tactic: 0 Time: 0.047708 [TRT] Fastest Tactic: 1002 Time: 0.0311195 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(27440,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4c/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.057735 [TRT] Tactic: 0 Time: 0.040469 [TRT] Fastest Tactic: 0 Time: 0.040469 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(27440,196:2,14,1) -> Half(50176,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4c/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.045052 [TRT] Tactic: 0 Time: 0.006745 [TRT] Fastest Tactic: 0 Time: 0.006745 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning format combination: Float(100352,196,14,1) -> Float(56448,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.831302 [TRT] Tactic: 655359 Time: 0.794271 [TRT] Tactic: 786431 Time: 0.910547 [TRT] Tactic: 851967 Time: 1.23734 [TRT] Tactic: 1179647 Time: 1.07281 [TRT] Tactic: 1310719 Time: 1.81591 [TRT] Tactic: 1376255 Time: 0.755287 [TRT] Tactic: 1441791 Time: 1.25703 [TRT] Tactic: 1507327 Time: 1.25042 [TRT] Tactic: 1638399 Time: 1.31495 [TRT] Tactic: 1835007 Time: 0.908854 [TRT] Tactic: 1900543 Time: 1.16445 [TRT] Tactic: 2097151 Time: 1.0213 [TRT] Tactic: 2162687 Time: 0.788802 [TRT] Tactic: 2293759 Time: 0.816589 [TRT] Tactic: 2359295 Time: 0.976667 [TRT] Tactic: 2686975 Time: 0.904453 [TRT] Tactic: 3080191 Time: 0.896016 [TRT] Tactic: 3342335 Time: 1.16107 [TRT] Tactic: 3407871 Time: 0.883776 [TRT] Tactic: 3538943 Time: 0.928828 [TRT] Tactic: 3670015 Time: 0.717839 [TRT] Tactic: 3932159 Time: 1.34943 [TRT] Tactic: 3997695 Time: 0.907135 [TRT] Tactic: 4063231 Time: 1.06044 [TRT] Tactic: 4194303 Time: 0.869167 [TRT] Tactic: 4259839 Time: 1.05021 [TRT] Tactic: 4325375 Time: 1.10661 [TRT] Tactic: 4521983 Time: 1.13104 [TRT] Tactic: 4587519 Time: 0.997187 [TRT] Tactic: 4653055 Time: 1.05638 [TRT] Tactic: 4915199 Time: 0.84526 [TRT] Tactic: 4980735 Time: 1.09445 [TRT] Tactic: 5177343 Time: 1.28682 [TRT] Tactic: 5242879 Time: 0.795391 [TRT] Tactic: 5373951 Time: 1.23398 [TRT] Tactic: 5439487 Time: 0.890182 [TRT] Tactic: 5570559 Time: 0.817578 [TRT] Tactic: 5636095 Time: 1.05859 [TRT] Tactic: 5701631 Time: 0.988047 [TRT] Tactic: 5767167 Time: 1.3494 [TRT] Tactic: 5832703 Time: 0.878281 [TRT] Tactic: 5898239 Time: 0.776536 [TRT] Tactic: 6029311 Time: 0.760937 [TRT] Tactic: 6225919 Time: 0.847813 [TRT] Tactic: 6291455 Time: 1.08247 [TRT] Tactic: 6422527 Time: 0.888698 [TRT] Tactic: 6750207 Time: 0.854766 [TRT] Tactic: 6815743 Time: 0.862553 [TRT] Tactic: 6946815 Time: 1.16422 [TRT] Tactic: 7012351 Time: 1.02013 [TRT] Tactic: 7077887 Time: 0.922969 [TRT] Tactic: 7143423 Time: 1.30055 [TRT] Tactic: 7208959 Time: 1.04346 [TRT] Tactic: 7340031 Time: 0.833854 [TRT] Tactic: 7405567 Time: 0.940885 [TRT] Tactic: 7536639 Time: 0.969973 [TRT] Tactic: 7602175 Time: 1.07938 [TRT] Tactic: 7733247 Time: 0.84875 [TRT] Tactic: 7798783 Time: 0.914531 [TRT] Tactic: 8191999 Time: 1.35039 [TRT] Tactic: 8257535 Time: 0.880651 [TRT] Tactic: 8323071 Time: 0.792943 [TRT] Tactic: 8650751 Time: 1.24068 [TRT] Tactic: 8716287 Time: 0.943646 [TRT] Tactic: 9109503 Time: 1.1037 [TRT] Tactic: 9568255 Time: 0.845026 [TRT] Tactic: 9895935 Time: 0.865233 [TRT] Tactic: 10223615 Time: 0.905807 [TRT] Tactic: 10354687 Time: 1.07164 [TRT] Tactic: 10551295 Time: 0.798125 [TRT] Tactic: 10747903 Time: 0.799296 [TRT] Tactic: 10944511 Time: 1.09307 [TRT] Fastest Tactic: 3670015 Time: 0.717839 [TRT] --------------- Timing Runner: inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce (CudnnConvolution) [TRT] Tactic: 0 Time: 0.869245 [TRT] Tactic: 1 Time: 0.733646 [TRT] Tactic: 2 Time: 1.03271 [TRT] Tactic: 4 skipped. Scratch requested: 341508096, available: 33554432 [TRT] Tactic: 5 Time: 10.5304 [TRT] Fastest Tactic: 1 Time: 0.733646 [TRT] --------------- Timing Runner: inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce (CaskConvolution) [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.501224 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v0 Tactic: 1698681053543049347 [TRT] Tactic: 1698681053543049347 Time: 0.49013 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.540156 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.45625 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.533437 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v1 Tactic: 5326823351883942011 [TRT] Tactic: 5326823351883942011 Time: 0.51599 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v0 Tactic: 5500448035057547314 [TRT] Tactic: 5500448035057547314 Time: 0.50526 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.468021 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.554245 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.544011 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v1 Tactic: -6576203419454146580 [TRT] Tactic: -6576203419454146580 Time: 0.465052 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.586614 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.491693 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.532969 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.516927 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.528073 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v0 Tactic: -377491875521947884 [TRT] Tactic: -377491875521947884 Time: 0.529401 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v1 Tactic: -37215280111360163 [TRT] Tactic: -37215280111360163 Time: 0.444427 [TRT] Fastest Tactic: -37215280111360163 Time: 0.444427 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -37215280111360163 [TRT] *************** Autotuning format combination: Float(100352,1,7168,512) -> Float(56448,1,4032,288) *************** [TRT] --------------- Timing Runner: inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce (CaskConvolution) [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 3886731678879822788 [TRT] Tactic: 3886731678879822788 Time: 0.45026 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 6629944304117643200 [TRT] Tactic: 6629944304117643200 Time: 0.647995 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.677994 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.449896 [TRT] Fastest Tactic: -7394439838318485025 Time: 0.449896 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(100352,196,14,1) -> Half(56448,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce (CudnnConvolution) [TRT] Tactic: 0 Time: 0.948984 [TRT] Tactic: 1 Time: 0.770182 [TRT] Tactic: 2 Time: 1.0132 [TRT] Tactic: 4 skipped. Scratch requested: 341508096, available: 33554432 [TRT] Tactic: 5 Time: 10.9341 [TRT] Fastest Tactic: 1 Time: 0.770182 [TRT] --------------- Timing Runner: inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 1 [TRT] *************** Autotuning format combination: Half(50176,196:2,14,1) -> Half(56448,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] *************** Autotuning format combination: Half(50176,196:2,14,1) -> Half(28224,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.403359 [TRT] Tactic: 655359 Time: 0.495495 [TRT] Tactic: 786431 Time: 0.578021 [TRT] Tactic: 851967 Time: 0.660339 [TRT] Tactic: 1179647 Time: 0.460364 [TRT] Tactic: 1310719 Time: 0.930886 [TRT] Tactic: 1376255 Time: 0.375338 [TRT] Tactic: 1441791 Time: 0.623074 [TRT] Tactic: 1507327 Time: 0.676355 [TRT] Tactic: 1638399 Time: 0.691979 [TRT] Tactic: 1835007 Time: 0.581303 [TRT] Tactic: 1900543 Time: 0.626901 [TRT] Tactic: 2097151 Time: 0.695417 [TRT] Tactic: 2162687 Time: 0.391537 [TRT] Tactic: 2293759 Time: 0.42289 [TRT] Tactic: 2359295 Time: 0.501094 [TRT] Tactic: 2686975 Time: 0.708698 [TRT] Tactic: 3080191 Time: 0.507057 [TRT] Tactic: 3342335 Time: 0.623385 [TRT] Tactic: 3407871 Time: 0.415704 [TRT] Tactic: 3538943 Time: 0.451927 [TRT] Tactic: 3670015 Time: 0.472265 [TRT] Tactic: 3932159 Time: 0.587474 [TRT] Tactic: 3997695 Time: 0.566745 [TRT] Tactic: 4063231 Time: 0.563229 [TRT] Tactic: 4194303 Time: 0.49276 [TRT] Tactic: 4259839 Time: 0.693385 [TRT] Tactic: 4325375 Time: 0.563359 [TRT] Tactic: 4521983 Time: 0.546067 [TRT] Tactic: 4587519 Time: 0.58112 [TRT] Tactic: 4653055 Time: 0.550964 [TRT] Tactic: 4915199 Time: 0.48875 [TRT] Tactic: 4980735 Time: 0.566094 [TRT] Tactic: 5177343 Time: 0.544583 [TRT] Tactic: 5242879 Time: 0.368308 [TRT] Tactic: 5373951 Time: 0.543307 [TRT] Tactic: 5439487 Time: 0.507578 [TRT] Tactic: 5570559 Time: 0.550079 [TRT] Tactic: 5636095 Time: 0.552604 [TRT] Tactic: 5701631 Time: 0.444166 [TRT] Tactic: 5767167 Time: 0.629349 [TRT] Tactic: 5832703 Time: 0.409557 [TRT] Tactic: 5898239 Time: 0.460156 [TRT] Tactic: 6029311 Time: 0.375964 [TRT] Tactic: 6225919 Time: 0.395182 [TRT] Tactic: 6291455 Time: 0.458594 [TRT] Tactic: 6422527 Time: 0.458542 [TRT] Tactic: 6750207 Time: 0.497188 [TRT] Tactic: 6815743 Time: 0.406771 [TRT] Tactic: 6946815 Time: 0.573204 [TRT] Tactic: 7012351 Time: 0.696979 [TRT] Tactic: 7077887 Time: 0.430208 [TRT] Tactic: 7143423 Time: 0.654062 [TRT] Tactic: 7208959 Time: 0.475963 [TRT] Tactic: 7340031 Time: 0.49 [TRT] Tactic: 7405567 Time: 0.468464 [TRT] Tactic: 7536639 Time: 0.492396 [TRT] Tactic: 7602175 Time: 0.519193 [TRT] Tactic: 7733247 Time: 0.464765 [TRT] Tactic: 7798783 Time: 0.578099 [TRT] Tactic: 8191999 Time: 0.686172 [TRT] Tactic: 8257535 Time: 0.503802 [TRT] Tactic: 8323071 Time: 0.461224 [TRT] Tactic: 8650751 Time: 0.605469 [TRT] Tactic: 8716287 Time: 0.450496 [TRT] Tactic: 9109503 Time: 0.693594 [TRT] Tactic: 9568255 Time: 0.489948 [TRT] Tactic: 9895935 Time: 0.490885 [TRT] Tactic: 10223615 Time: 0.708385 [TRT] Tactic: 10354687 Time: 0.693125 [TRT] Tactic: 10551295 Time: 0.404687 [TRT] Tactic: 10747903 Time: 0.409141 [TRT] Tactic: 10944511 Time: 0.564584 [TRT] Fastest Tactic: 5242879 Time: 0.368308 [TRT] --------------- Timing Runner: inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce (CaskConvolution) [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_interior_nn_v1 Tactic: 3066127711859985668 [TRT] Tactic: 3066127711859985668 Time: 0.245573 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.263359 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.260026 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.245261 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] Tactic: 8163473458334948789 Time: 0.233411 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.277604 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.27737 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.237057 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_interior_nn_v1 Tactic: -1716393687483585322 [TRT] Tactic: -1716393687483585322 Time: 0.272917 [TRT] Fastest Tactic: 8163473458334948789 Time: 0.233411 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 8163473458334948789 [TRT] *************** Autotuning Reformat:Float(56448,196,14,1) -> Float(56448,1,4032,288) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.086849 [TRT] Tactic: 0 Time: 0.0910675 [TRT] Fastest Tactic: 1002 Time: 0.086849 [TRT] *************** Autotuning Reformat:Float(56448,196,14,1) -> Half(56448,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.52086 [TRT] Tactic: 0 Time: 0.0657815 [TRT] Fastest Tactic: 0 Time: 0.0657815 [TRT] *************** Autotuning Reformat:Float(56448,196,14,1) -> Half(28224,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0956515 [TRT] Tactic: 0 Time: 0.0503385 [TRT] Fastest Tactic: 0 Time: 0.0503385 [TRT] *************** Autotuning Reformat:Float(56448,1,4032,288) -> Float(56448,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.108463 [TRT] Tactic: 0 Time: 0.084974 [TRT] Fastest Tactic: 0 Time: 0.084974 [TRT] *************** Autotuning Reformat:Float(56448,1,4032,288) -> Half(56448,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.069401 [TRT] Tactic: 0 Time: 0.0845575 [TRT] Fastest Tactic: 1002 Time: 0.069401 [TRT] *************** Autotuning Reformat:Float(56448,1,4032,288) -> Half(28224,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.082604 [TRT] Tactic: 0 Time: 0.0986455 [TRT] Fastest Tactic: 1002 Time: 0.082604 [TRT] *************** Autotuning Reformat:Half(56448,196,14,1) -> Float(56448,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.56786 [TRT] Tactic: 0 Time: 0.0539585 [TRT] Fastest Tactic: 0 Time: 0.0539585 [TRT] *************** Autotuning Reformat:Half(56448,196,14,1) -> Float(56448,1,4032,288) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0626305 [TRT] Tactic: 0 Time: 0.0901045 [TRT] Fastest Tactic: 1002 Time: 0.0626305 [TRT] *************** Autotuning Reformat:Half(56448,196,14,1) -> Half(28224,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0671095 [TRT] Tactic: 0 Time: 0.0497395 [TRT] Fastest Tactic: 0 Time: 0.0497395 [TRT] *************** Autotuning Reformat:Half(28224,196:2,14,1) -> Float(56448,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0911195 [TRT] Tactic: 0 Time: 0.0447135 [TRT] Fastest Tactic: 0 Time: 0.0447135 [TRT] *************** Autotuning Reformat:Half(28224,196:2,14,1) -> Float(56448,1,4032,288) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.063672 [TRT] Tactic: 0 Time: 0.104427 [TRT] Fastest Tactic: 1002 Time: 0.063672 [TRT] *************** Autotuning Reformat:Half(28224,196:2,14,1) -> Half(56448,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.126354 [TRT] Tactic: 0 Time: 0.0428905 [TRT] Fastest Tactic: 0 Time: 0.0428905 [TRT] *************** Autotuning Reformat:Float(56448,196,14,1) -> Float(56448,1,4032,288) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.05276 [TRT] Tactic: 0 Time: 0.045938 [TRT] Fastest Tactic: 0 Time: 0.045938 [TRT] *************** Autotuning Reformat:Float(56448,196,14,1) -> Half(56448,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.92073 [TRT] Tactic: 0 Time: 0.042578 [TRT] Fastest Tactic: 0 Time: 0.042578 [TRT] *************** Autotuning Reformat:Float(56448,196,14,1) -> Half(28224,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0520055 [TRT] Tactic: 0 Time: 0.0509375 [TRT] Fastest Tactic: 0 Time: 0.0509375 [TRT] *************** Autotuning Reformat:Float(56448,1,4032,288) -> Float(56448,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0626565 [TRT] Tactic: 0 Time: 0.042734 [TRT] Fastest Tactic: 0 Time: 0.042734 [TRT] *************** Autotuning Reformat:Float(56448,1,4032,288) -> Half(56448,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.039792 [TRT] Tactic: 0 Time: 0.04263 [TRT] Fastest Tactic: 1002 Time: 0.039792 [TRT] *************** Autotuning Reformat:Float(56448,1,4032,288) -> Half(28224,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0450785 [TRT] Tactic: 0 Time: 0.051432 [TRT] Fastest Tactic: 1002 Time: 0.0450785 [TRT] *************** Autotuning Reformat:Half(56448,196,14,1) -> Float(56448,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.942865 [TRT] Tactic: 0 Time: 0.028672 [TRT] Fastest Tactic: 0 Time: 0.028672 [TRT] *************** Autotuning Reformat:Half(56448,196,14,1) -> Float(56448,1,4032,288) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.038203 [TRT] Tactic: 0 Time: 0.0472395 [TRT] Fastest Tactic: 1002 Time: 0.038203 [TRT] *************** Autotuning Reformat:Half(56448,196,14,1) -> Half(28224,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.038177 [TRT] Tactic: 0 Time: 0.0264845 [TRT] Fastest Tactic: 0 Time: 0.0264845 [TRT] *************** Autotuning Reformat:Half(28224,196:2,14,1) -> Float(56448,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.049089 [TRT] Tactic: 0 Time: 0.0239065 [TRT] Fastest Tactic: 0 Time: 0.0239065 [TRT] *************** Autotuning Reformat:Half(28224,196:2,14,1) -> Float(56448,1,4032,288) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.038099 [TRT] Tactic: 0 Time: 0.0542705 [TRT] Fastest Tactic: 1002 Time: 0.038099 [TRT] *************** Autotuning Reformat:Half(28224,196:2,14,1) -> Half(56448,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.074062 [TRT] Tactic: 0 Time: 0.045052 [TRT] Fastest Tactic: 0 Time: 0.045052 [TRT] *************** Autotuning format combination: Float(56448,196,14,1) -> Float(103488,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/3x3 + inception_4d/relu_3x3 (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4d/3x3 + inception_4d/relu_3x3 (FusedConvActConvolution) [TRT] Tactic: 524287 Time: 1.03378 [TRT] Tactic: 720895 Time: 1.25471 [TRT] Tactic: 983039 Time: 1.0619 [TRT] Tactic: 1048575 Time: 1.05495 [TRT] Tactic: 1703935 Time: 1.03531 [TRT] Tactic: 1769471 Time: 1.06664 [TRT] Tactic: 1966079 Time: 1.16732 [TRT] Tactic: 2031615 Time: 1.12617 [TRT] Tactic: 2228223 Time: 1.33422 [TRT] Tactic: 2424831 Time: 1.46745 [TRT] Tactic: 2621439 Time: 1.23854 [TRT] Tactic: 2752511 Time: 1.03065 [TRT] Tactic: 2818047 Time: 1.67477 [TRT] Tactic: 2883583 Time: 1.32747 [TRT] Tactic: 3014655 Time: 0.963854 [TRT] Tactic: 3145727 Time: 1.00302 [TRT] Tactic: 3473407 Time: 1.47956 [TRT] Tactic: 3604479 Time: 0.956198 [TRT] Tactic: 3735551 Time: 1.66122 [TRT] Tactic: 4390911 Time: 1.0093 [TRT] Tactic: 5046271 Time: 1.0062 [TRT] Tactic: 5963775 Time: 1.06076 [TRT] Tactic: 6160383 Time: 1.00729 [TRT] Tactic: 6488063 Time: 0.892812 [TRT] Tactic: 6881279 Time: 1.06112 [TRT] Tactic: 7274495 Time: 1.3968 [TRT] Tactic: 7864319 Time: 1.30128 [TRT] Tactic: 7995391 Time: 1.24234 [TRT] Tactic: 8585215 Time: 0.943047 [TRT] Tactic: 8847359 Time: 0.978932 [TRT] Tactic: 8978431 Time: 1.06242 [TRT] Tactic: 9043967 Time: 0.881615 [TRT] Tactic: 9175039 Time: 0.953567 [TRT] Tactic: 9502719 Time: 1.0625 [TRT] Tactic: 9830399 Time: 1.26867 [TRT] Tactic: 9961471 Time: 1.08175 [TRT] Tactic: 10027007 Time: 0.904557 [TRT] Tactic: 10092543 Time: 1.01 [TRT] Tactic: 10289151 Time: 1.17068 [TRT] Tactic: 10485759 Time: 0.910547 [TRT] Tactic: 10682367 Time: 1.25674 [TRT] Tactic: 10813439 Time: 1.21224 [TRT] Fastest Tactic: 9043967 Time: 0.881615 [TRT] --------------- Timing Runner: inception_4d/3x3 + inception_4d/relu_3x3 (CudnnConvolution) [TRT] Tactic: 0 Time: 2.3162 [TRT] Tactic: 1 Time: 1.23737 [TRT] Tactic: 2 Time: 2.07781 [TRT] Tactic: 4 skipped. Scratch requested: 97376256, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 182366208, available: 33554432 [TRT] Tactic: 6 Time: 0.919167 [TRT] Fastest Tactic: 6 Time: 0.919167 [TRT] --------------- Timing Runner: inception_4d/3x3 + inception_4d/relu_3x3 (CaskConvolution) [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 1.24815 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v0 Tactic: 1754984623894446479 [TRT] Tactic: 1754984623894446479 Time: 1.50305 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v0 Tactic: 3611739942397549984 [TRT] Tactic: 3611739942397549984 Time: 1.3174 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148n_nt_v1 Tactic: 3827454225649558724 [TRT] Tactic: 3827454225649558724 Time: 0.827057 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v1 Tactic: 4337000649858996379 [TRT] Tactic: 4337000649858996379 Time: 1.16503 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 1.32164 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 1.08096 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 1.28201 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 5921334924264294896 [TRT] Tactic: 5921334924264294896 Time: 0.632447 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 1.14427 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 1.27888 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148t_nt_v1 Tactic: 7852627285308570038 [TRT] Tactic: 7852627285308570038 Time: 0.793932 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v1 Tactic: -9137461792520977713 [TRT] Tactic: -9137461792520977713 Time: 1.32896 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148n_nt_v0 Tactic: -8776506421218919509 [TRT] Tactic: -8776506421218919509 Time: 0.780234 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 1.32669 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v0 Tactic: -8133971918129952780 [TRT] Tactic: -8133971918129952780 Time: 1.27703 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v1 Tactic: -6092040395344634144 [TRT] Tactic: -6092040395344634144 Time: 1.2868 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 1.50107 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 1.11175 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148t_nt_v0 Tactic: -2318106587342035239 [TRT] Tactic: -2318106587342035239 Time: 0.754869 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_mobile_relu_tile148t_nt_v0 Tactic: -1343271414618805657 [TRT] Tactic: -1343271414618805657 Time: 0.563203 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 1.26943 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 1.21367 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 1.2669 [TRT] Fastest Tactic: -1343271414618805657 Time: 0.563203 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -1343271414618805657 [TRT] *************** Autotuning format combination: Float(56448,1,4032,288) -> Float(103488,1,7392,528) *************** [TRT] --------------- Timing Runner: inception_4d/3x3 + inception_4d/relu_3x3 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4d/3x3 + inception_4d/relu_3x3 (CaskConvolution) [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 1.33661 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 1.06292 [TRT] Fastest Tactic: -7394439838318485025 Time: 1.06292 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(56448,196,14,1) -> Half(103488,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/3x3 + inception_4d/relu_3x3 (CudnnConvolution) [TRT] Tactic: 0 Time: 2.31122 [TRT] Tactic: 1 Time: 2.34339 [TRT] Tactic: 2 Time: 2.02398 [TRT] Tactic: 4 skipped. Scratch requested: 97376256, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 182366208, available: 33554432 [TRT] Tactic: 6 Time: 2.42164 [TRT] Fastest Tactic: 2 Time: 2.02398 [TRT] --------------- Timing Runner: inception_4d/3x3 + inception_4d/relu_3x3 (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 2 [TRT] *************** Autotuning format combination: Half(28224,196:2,14,1) -> Half(51744,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/3x3 + inception_4d/relu_3x3 (FusedConvActConvolution) [TRT] Tactic: 524287 Time: 0.542292 [TRT] Tactic: 720895 Time: 0.745625 [TRT] Tactic: 983039 Time: 0.58961 [TRT] Tactic: 1048575 Time: 0.583073 [TRT] Tactic: 1703935 Time: 0.590469 [TRT] Tactic: 1769471 Time: 3.75885 [TRT] Tactic: 1966079 Time: 0.721328 [TRT] Tactic: 2031615 Time: 0.597969 [TRT] Tactic: 2228223 Time: 0.737214 [TRT] Tactic: 2621439 Time: 0.677969 [TRT] Tactic: 2752511 Time: 0.567162 [TRT] Tactic: 2818047 Time: 0.922396 [TRT] Tactic: 2883583 Time: 0.839895 [TRT] Tactic: 3014655 Time: 0.552552 [TRT] Tactic: 3145727 Time: 0.584296 [TRT] Tactic: 3473407 Time: 0.869401 [TRT] Tactic: 3604479 Time: 0.544323 [TRT] Tactic: 3735551 Time: 0.839297 [TRT] Tactic: 4390911 Time: 0.618881 [TRT] Tactic: 5046271 Time: 0.546614 [TRT] Tactic: 5963775 Time: 0.576172 [TRT] Tactic: 6160383 Time: 0.576016 [TRT] Tactic: 6488063 Time: 0.500781 [TRT] Tactic: 6881279 Time: 0.562813 [TRT] Tactic: 7274495 Time: 0.775807 [TRT] Tactic: 7864319 Time: 0.724896 [TRT] Tactic: 7995391 Time: 0.702812 [TRT] Tactic: 8585215 Time: 0.528933 [TRT] Tactic: 8847359 Time: 0.535521 [TRT] Tactic: 8978431 Time: 0.609662 [TRT] Tactic: 9043967 Time: 0.487058 [TRT] Tactic: 9175039 Time: 0.546484 [TRT] Tactic: 9502719 Time: 0.607656 [TRT] Tactic: 9830399 Time: 0.667865 [TRT] Tactic: 10027007 Time: 0.500026 [TRT] Tactic: 10092543 Time: 0.593567 [TRT] Tactic: 10289151 Time: 0.718776 [TRT] Tactic: 10485759 Time: 0.497032 [TRT] Tactic: 10682367 Time: 0.667865 [TRT] Tactic: 10813439 Time: 0.726693 [TRT] Fastest Tactic: 9043967 Time: 0.487058 [TRT] --------------- Timing Runner: inception_4d/3x3 + inception_4d/relu_3x3 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4d/3x3 + inception_4d/relu_3x3 (CaskConvolution) [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.668516 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_large_nn_v1 Tactic: 3650389455493082349 [TRT] Tactic: 3650389455493082349 Time: 0.699791 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_winograd_fp16x2_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 4772821744921268633 [TRT] Tactic: 4772821744921268633 Time: 0.332787 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.562813 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.584401 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_large_nn_v1 Tactic: -6490690591794140522 [TRT] Tactic: -6490690591794140522 Time: 0.586875 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_large_nn_v1 Tactic: -4686027666808657977 [TRT] Tactic: -4686027666808657977 Time: 0.674141 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.643489 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.669818 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.553802 [TRT] Fastest Tactic: 4772821744921268633 Time: 0.332787 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 4772821744921268633 [TRT] *************** Autotuning Reformat:Float(103488,196,14,1) -> Float(103488,1,7392,528) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0883075 [TRT] Tactic: 0 Time: 0.090443 [TRT] Fastest Tactic: 1002 Time: 0.0883075 [TRT] *************** Autotuning Reformat:Float(103488,196,14,1) -> Half(103488,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.83505 [TRT] Tactic: 0 Time: 0.0827605 [TRT] Fastest Tactic: 0 Time: 0.0827605 [TRT] *************** Autotuning Reformat:Float(103488,196,14,1) -> Half(51744,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0955725 [TRT] Tactic: 0 Time: 0.050729 [TRT] Fastest Tactic: 0 Time: 0.050729 [TRT] *************** Autotuning Reformat:Float(103488,1,7392,528) -> Float(103488,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.108125 [TRT] Tactic: 0 Time: 0.0857295 [TRT] Fastest Tactic: 0 Time: 0.0857295 [TRT] *************** Autotuning Reformat:Float(103488,1,7392,528) -> Half(103488,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0691665 [TRT] Tactic: 0 Time: 0.083516 [TRT] Fastest Tactic: 1002 Time: 0.0691665 [TRT] *************** Autotuning Reformat:Float(103488,1,7392,528) -> Half(51744,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.083151 [TRT] Tactic: 0 Time: 0.099323 [TRT] Fastest Tactic: 1002 Time: 0.083151 [TRT] *************** Autotuning Reformat:Half(103488,196,14,1) -> Float(103488,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.87682 [TRT] Tactic: 0 Time: 0.0845055 [TRT] Fastest Tactic: 0 Time: 0.0845055 [TRT] *************** Autotuning Reformat:Half(103488,196,14,1) -> Float(103488,1,7392,528) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.063411 [TRT] Tactic: 0 Time: 0.0896355 [TRT] Fastest Tactic: 1002 Time: 0.063411 [TRT] *************** Autotuning Reformat:Half(103488,196,14,1) -> Half(51744,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.068724 [TRT] Tactic: 0 Time: 0.0495305 [TRT] Fastest Tactic: 0 Time: 0.0495305 [TRT] *************** Autotuning Reformat:Half(51744,196:2,14,1) -> Float(103488,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.090885 [TRT] Tactic: 0 Time: 0.0969015 [TRT] Fastest Tactic: 1002 Time: 0.090885 [TRT] *************** Autotuning Reformat:Half(51744,196:2,14,1) -> Float(103488,1,7392,528) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0641145 [TRT] Tactic: 0 Time: 0.104141 [TRT] Fastest Tactic: 1002 Time: 0.0641145 [TRT] *************** Autotuning Reformat:Half(51744,196:2,14,1) -> Half(103488,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.127578 [TRT] Tactic: 0 Time: 0.087396 [TRT] Fastest Tactic: 0 Time: 0.087396 [TRT] *************** Autotuning Reformat:Float(56448,196,14,1) -> Float(56448,1,4032,288) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0241145 [TRT] Tactic: 0 Time: 0.0127345 [TRT] Fastest Tactic: 0 Time: 0.0127345 [TRT] *************** Autotuning Reformat:Float(56448,196,14,1) -> Half(56448,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.209453 [TRT] Tactic: 0 Time: 0.0125 [TRT] Fastest Tactic: 0 Time: 0.0125 [TRT] *************** Autotuning Reformat:Float(56448,196,14,1) -> Half(28224,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.015052 [TRT] Tactic: 0 Time: 0.0148955 [TRT] Fastest Tactic: 0 Time: 0.0148955 [TRT] *************** Autotuning Reformat:Float(56448,1,4032,288) -> Float(56448,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0266925 [TRT] Tactic: 0 Time: 0.012734 [TRT] Fastest Tactic: 0 Time: 0.012734 [TRT] *************** Autotuning Reformat:Float(56448,1,4032,288) -> Half(56448,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.012813 [TRT] Tactic: 0 Time: 0.012708 [TRT] Fastest Tactic: 0 Time: 0.012708 [TRT] *************** Autotuning Reformat:Float(56448,1,4032,288) -> Half(28224,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.015313 [TRT] Tactic: 0 Time: 0.014843 [TRT] Fastest Tactic: 0 Time: 0.014843 [TRT] *************** Autotuning Reformat:Half(56448,196,14,1) -> Float(56448,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.215364 [TRT] Tactic: 0 Time: 0.010156 [TRT] Fastest Tactic: 0 Time: 0.010156 [TRT] *************** Autotuning Reformat:Half(56448,196,14,1) -> Float(56448,1,4032,288) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.012552 [TRT] Tactic: 0 Time: 0.0127345 [TRT] Fastest Tactic: 1002 Time: 0.012552 [TRT] *************** Autotuning Reformat:Half(56448,196,14,1) -> Half(28224,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0148175 [TRT] Tactic: 0 Time: 0.0078645 [TRT] Fastest Tactic: 0 Time: 0.0078645 [TRT] *************** Autotuning Reformat:Half(28224,196:2,14,1) -> Float(56448,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.016354 [TRT] Tactic: 0 Time: 0.0078905 [TRT] Fastest Tactic: 0 Time: 0.0078905 [TRT] *************** Autotuning Reformat:Half(28224,196:2,14,1) -> Float(56448,1,4032,288) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0103645 [TRT] Tactic: 0 Time: 0.0148435 [TRT] Fastest Tactic: 1002 Time: 0.0103645 [TRT] *************** Autotuning Reformat:Half(28224,196:2,14,1) -> Half(56448,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.025234 [TRT] Tactic: 0 Time: 0.012526 [TRT] Fastest Tactic: 0 Time: 0.012526 [TRT] *************** Autotuning format combination: Float(56448,196,14,1) -> Float(103488,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/5x5 + inception_4d/relu_5x5 (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4d/5x5 + inception_4d/relu_5x5 (FusedConvActConvolution) [TRT] Tactic: 393215 Time: 0.177709 [TRT] Tactic: 917503 Time: 0.156849 [TRT] Tactic: 1114111 Time: 0.166849 [TRT] Tactic: 1245183 Time: 0.172057 [TRT] Tactic: 1572863 Time: 0.175756 [TRT] Tactic: 2490367 Time: 0.190131 [TRT] Tactic: 2555903 Time: 0.176979 [TRT] Tactic: 2949119 Time: 0.147839 [TRT] Tactic: 3211263 Time: 0.303567 [TRT] Tactic: 3801087 Time: 0.161042 [TRT] Tactic: 3866623 Time: 0.16362 [TRT] Tactic: 4128767 Time: 0.145052 [TRT] Tactic: 4456447 Time: 0.150104 [TRT] Tactic: 4718591 Time: 0.148438 [TRT] Tactic: 4784127 Time: 0.397812 [TRT] Tactic: 4849663 Time: 0.156068 [TRT] Tactic: 5111807 Time: 0.150755 [TRT] Tactic: 5308415 Time: 0.181823 [TRT] Tactic: 5505023 Time: 0.266614 [TRT] Tactic: 6094847 Time: 0.166954 [TRT] Tactic: 6356991 Time: 0.176798 [TRT] Tactic: 6553599 Time: 0.177266 [TRT] Tactic: 6619135 Time: 0.194375 [TRT] Tactic: 6684671 Time: 0.281042 [TRT] Tactic: 7471103 Time: 0.165547 [TRT] Tactic: 7667711 Time: 0.148151 [TRT] Tactic: 7929855 Time: 0.161979 [TRT] Tactic: 8060927 Time: 0.180052 [TRT] Tactic: 8126463 Time: 0.158412 [TRT] Tactic: 8388607 Time: 0.171484 [TRT] Tactic: 8519679 Time: 0.197682 [TRT] Tactic: 8781823 Time: 0.189375 [TRT] Tactic: 8912895 Time: 0.15638 [TRT] Tactic: 9240575 Time: 0.147396 [TRT] Tactic: 9306111 Time: 0.248594 [TRT] Tactic: 9371647 Time: 0.161172 [TRT] Tactic: 9437183 Time: 0.146875 [TRT] Tactic: 9633791 Time: 0.153177 [TRT] Tactic: 9699327 Time: 0.171328 [TRT] Tactic: 9764863 Time: 0.183854 [TRT] Tactic: 10158079 Time: 0.176953 [TRT] Tactic: 10420223 Time: 0.160651 [TRT] Tactic: 10616831 Time: 0.160287 [TRT] Tactic: 10878975 Time: 0.193307 [TRT] Fastest Tactic: 4128767 Time: 0.145052 [TRT] --------------- Timing Runner: inception_4d/5x5 + inception_4d/relu_5x5 (CudnnConvolution) [TRT] Tactic: 0 Time: 0.331979 [TRT] Tactic: 1 Time: 0.332995 [TRT] Tactic: 2 Time: 0.550676 [TRT] Tactic: 4 Time: 2.89615 [TRT] Tactic: 5 Time: 2.8862 [TRT] Fastest Tactic: 0 Time: 0.331979 [TRT] --------------- Timing Runner: inception_4d/5x5 + inception_4d/relu_5x5 (CaskConvolution) [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.246875 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v0 Tactic: 1754984623894446479 [TRT] Tactic: 1754984623894446479 Time: 0.24875 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v0 Tactic: 3611739942397549984 [TRT] Tactic: 3611739942397549984 Time: 0.294844 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v1 Tactic: 4337000649858996379 [TRT] Tactic: 4337000649858996379 Time: 0.199322 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.2975 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.157214 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.277448 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.19474 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.189401 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v1 Tactic: -9137461792520977713 [TRT] Tactic: -9137461792520977713 Time: 0.297994 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.294452 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v0 Tactic: -8133971918129952780 [TRT] Tactic: -8133971918129952780 Time: 0.221432 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v1 Tactic: -6092040395344634144 [TRT] Tactic: -6092040395344634144 Time: 0.254322 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.246146 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.193542 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.218881 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.193594 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.273333 [TRT] Fastest Tactic: 5137655947464784826 Time: 0.157214 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 4128767 [TRT] *************** Autotuning format combination: Float(56448,1,4032,288) -> Float(103488,1,7392,528) *************** [TRT] --------------- Timing Runner: inception_4d/5x5 + inception_4d/relu_5x5 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4d/5x5 + inception_4d/relu_5x5 (CaskConvolution) [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.177422 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.14112 [TRT] Fastest Tactic: -7394439838318485025 Time: 0.14112 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(56448,196,14,1) -> Half(103488,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/5x5 + inception_4d/relu_5x5 (CudnnConvolution) [TRT] Tactic: 0 Time: 0.375495 [TRT] Tactic: 1 Time: 0.355781 [TRT] Tactic: 2 Time: 0.548724 [TRT] Tactic: 4 Time: 2.90068 [TRT] Tactic: 5 Time: 2.90154 [TRT] Fastest Tactic: 1 Time: 0.355781 [TRT] --------------- Timing Runner: inception_4d/5x5 + inception_4d/relu_5x5 (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 1 [TRT] *************** Autotuning format combination: Half(28224,196:2,14,1) -> Half(51744,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/5x5 + inception_4d/relu_5x5 (FusedConvActConvolution) [TRT] Tactic: 393215 Time: 0.0975785 [TRT] Tactic: 917503 Time: 0.093516 [TRT] Tactic: 1114111 Time: 0.087943 [TRT] Tactic: 1245183 Time: 0.0929425 [TRT] Tactic: 1572863 Time: 0.100782 [TRT] Tactic: 2490367 Time: 0.121718 [TRT] Tactic: 2555903 Time: 0.119584 [TRT] Tactic: 2949119 Time: 0.0830465 [TRT] Tactic: 3211263 Time: 0.169636 [TRT] Tactic: 3801087 Time: 0.0908595 [TRT] Tactic: 3866623 Time: 0.0918225 [TRT] Tactic: 4128767 Time: 0.0825 [TRT] Tactic: 4456447 Time: 0.087526 [TRT] Tactic: 4718591 Time: 0.0806775 [TRT] Tactic: 4784127 Time: 0.198229 [TRT] Tactic: 4849663 Time: 0.0840885 [TRT] Tactic: 5111807 Time: 0.084323 [TRT] Tactic: 5308415 Time: 0.100573 [TRT] Tactic: 5505023 Time: 0.151588 [TRT] Tactic: 6094847 Time: 0.0939065 [TRT] Tactic: 6356991 Time: 0.116822 [TRT] Tactic: 6553599 Time: 0.098021 [TRT] Tactic: 6619135 Time: 0.100678 [TRT] Tactic: 6684671 Time: 0.161953 [TRT] Tactic: 7471103 Time: 0.0973955 [TRT] Tactic: 7667711 Time: 0.07961 [TRT] Tactic: 7929855 Time: 0.084167 [TRT] Tactic: 8060927 Time: 0.105651 [TRT] Tactic: 8126463 Time: 0.0813535 [TRT] Tactic: 8388607 Time: 0.103203 [TRT] Tactic: 8519679 Time: 0.103411 [TRT] Tactic: 8781823 Time: 0.09513 [TRT] Tactic: 8912895 Time: 0.0899225 [TRT] Tactic: 9240575 Time: 0.0756775 [TRT] Tactic: 9306111 Time: 0.127317 [TRT] Tactic: 9371647 Time: 0.085677 [TRT] Tactic: 9437183 Time: 0.083802 [TRT] Tactic: 9633791 Time: 0.083932 [TRT] Tactic: 9699327 Time: 0.098541 [TRT] Tactic: 9764863 Time: 0.106849 [TRT] Tactic: 10158079 Time: 0.10112 [TRT] Tactic: 10420223 Time: 0.094115 [TRT] Tactic: 10616831 Time: 0.09125 [TRT] Tactic: 10878975 Time: 0.110182 [TRT] Fastest Tactic: 9240575 Time: 0.0756775 [TRT] --------------- Timing Runner: inception_4d/5x5 + inception_4d/relu_5x5 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4d/5x5 + inception_4d/relu_5x5 (CaskConvolution) [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.13224 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_large_nn_v1 Tactic: 3650389455493082349 [TRT] Tactic: 3650389455493082349 Time: 0.133281 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.099922 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.104272 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_large_nn_v1 Tactic: -6490690591794140522 [TRT] Tactic: -6490690591794140522 Time: 0.105834 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_large_nn_v1 Tactic: -4686027666808657977 [TRT] Tactic: -4686027666808657977 Time: 0.15336 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.140182 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.151719 [TRT] inception_4d/5x5 + inception_4d/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.0845315 [TRT] Fastest Tactic: -2409163523992614473 Time: 0.0845315 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 9240575 [TRT] *************** Autotuning Reformat:Float(103488,196,14,1) -> Float(103488,1,7392,528) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0230465 [TRT] Tactic: 0 Time: 0.021901 [TRT] Fastest Tactic: 0 Time: 0.021901 [TRT] *************** Autotuning Reformat:Float(103488,196,14,1) -> Half(103488,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.413229 [TRT] Tactic: 0 Time: 0.0217445 [TRT] Fastest Tactic: 0 Time: 0.0217445 [TRT] *************** Autotuning Reformat:Float(103488,196,14,1) -> Half(51744,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.026562 [TRT] Tactic: 0 Time: 0.014896 [TRT] Fastest Tactic: 0 Time: 0.014896 [TRT] *************** Autotuning Reformat:Float(103488,1,7392,528) -> Float(103488,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0311455 [TRT] Tactic: 0 Time: 0.0211985 [TRT] Fastest Tactic: 0 Time: 0.0211985 [TRT] *************** Autotuning Reformat:Float(103488,1,7392,528) -> Half(103488,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.01987 [TRT] Tactic: 0 Time: 0.0217445 [TRT] Fastest Tactic: 1002 Time: 0.01987 [TRT] *************** Autotuning Reformat:Float(103488,1,7392,528) -> Half(51744,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.024245 [TRT] Tactic: 0 Time: 0.0242445 [TRT] Fastest Tactic: 0 Time: 0.0242445 [TRT] *************** Autotuning Reformat:Half(103488,196,14,1) -> Float(103488,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.422786 [TRT] Tactic: 0 Time: 0.021693 [TRT] Fastest Tactic: 0 Time: 0.021693 [TRT] *************** Autotuning Reformat:Half(103488,196,14,1) -> Float(103488,1,7392,528) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.017292 [TRT] Tactic: 0 Time: 0.023177 [TRT] Fastest Tactic: 1002 Time: 0.017292 [TRT] *************** Autotuning Reformat:Half(103488,196,14,1) -> Half(51744,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.023125 [TRT] Tactic: 0 Time: 0.013776 [TRT] Fastest Tactic: 0 Time: 0.013776 [TRT] *************** Autotuning Reformat:Half(51744,196:2,14,1) -> Float(103488,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.026537 [TRT] Tactic: 0 Time: 0.0241405 [TRT] Fastest Tactic: 0 Time: 0.0241405 [TRT] *************** Autotuning Reformat:Half(51744,196:2,14,1) -> Float(103488,1,7392,528) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.017265 [TRT] Tactic: 0 Time: 0.026563 [TRT] Fastest Tactic: 1002 Time: 0.017265 [TRT] *************** Autotuning Reformat:Half(51744,196:2,14,1) -> Half(103488,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0334895 [TRT] Tactic: 0 Time: 0.021849 [TRT] Fastest Tactic: 0 Time: 0.021849 [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning format combination: Float(100352,196,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning format combination: Half(100352,196,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning format combination: Half(50176,196:2,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(100352,1,7168,512) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Half(100352,196,14,1) -> Half(50176,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Float(100352,1,7168,512) *************** [TRT] *************** Autotuning Reformat:Half(50176,196:2,14,1) -> Half(100352,196,14,1) *************** [TRT] *************** Autotuning format combination: Float(100352,196,14,1) -> Float(103488,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/pool_proj + inception_4d/relu_pool_proj (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4d/pool_proj + inception_4d/relu_pool_proj (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.210677 [TRT] Tactic: 655359 Time: 0.235234 [TRT] Tactic: 786431 Time: 0.199218 [TRT] Tactic: 851967 Time: 0.337031 [TRT] Tactic: 1179647 Time: 0.188307 [TRT] Tactic: 1310719 Time: 0.413255 [TRT] Tactic: 1376255 Time: 0.221771 [TRT] Tactic: 1441791 Time: 0.240157 [TRT] Tactic: 1507327 Time: 0.440131 [TRT] Tactic: 1638399 Time: 0.280755 [TRT] Tactic: 1835007 Time: 0.209167 [TRT] Tactic: 1900543 Time: 0.421302 [TRT] Tactic: 2097151 Time: 0.226692 [TRT] Tactic: 2162687 Time: 0.234583 [TRT] Tactic: 2293759 Time: 0.222214 [TRT] Tactic: 2359295 Time: 0.247682 [TRT] Tactic: 2686975 Time: 0.218151 [TRT] Tactic: 3080191 Time: 0.275286 [TRT] Tactic: 3342335 Time: 0.348307 [TRT] Tactic: 3407871 Time: 0.167213 [TRT] Tactic: 3538943 Time: 0.164167 [TRT] Tactic: 3670015 Time: 0.276458 [TRT] Tactic: 3932159 Time: 0.380443 [TRT] Tactic: 3997695 Time: 0.20698 [TRT] Tactic: 4063231 Time: 0.293464 [TRT] Tactic: 4194303 Time: 0.202396 [TRT] Tactic: 4259839 Time: 0.235469 [TRT] Tactic: 4325375 Time: 0.257787 [TRT] Tactic: 4521983 Time: 0.255547 [TRT] Tactic: 4587519 Time: 0.225859 [TRT] Tactic: 4653055 Time: 0.205469 [TRT] Tactic: 4915199 Time: 0.210443 [TRT] Tactic: 4980735 Time: 0.303907 [TRT] Tactic: 5177343 Time: 0.223359 [TRT] Tactic: 5242879 Time: 0.14987 [TRT] Tactic: 5373951 Time: 0.213021 [TRT] Tactic: 5439487 Time: 0.197656 [TRT] Tactic: 5570559 Time: 0.22888 [TRT] Tactic: 5636095 Time: 0.296692 [TRT] Tactic: 5701631 Time: 0.227057 [TRT] Tactic: 5767167 Time: 0.279349 [TRT] Tactic: 5832703 Time: 0.166432 [TRT] Tactic: 5898239 Time: 0.154505 [TRT] Tactic: 6029311 Time: 0.214036 [TRT] Tactic: 6225919 Time: 0.146979 [TRT] Tactic: 6291455 Time: 0.187474 [TRT] Tactic: 6422527 Time: 0.287605 [TRT] Tactic: 6750207 Time: 0.19849 [TRT] Tactic: 6815743 Time: 0.150235 [TRT] Tactic: 6946815 Time: 0.28224 [TRT] Tactic: 7012351 Time: 0.225469 [TRT] Tactic: 7077887 Time: 0.161563 [TRT] Tactic: 7143423 Time: 0.292656 [TRT] Tactic: 7208959 Time: 0.227605 [TRT] Tactic: 7340031 Time: 0.161822 [TRT] Tactic: 7405567 Time: 0.187292 [TRT] Tactic: 7536639 Time: 0.176041 [TRT] Tactic: 7602175 Time: 0.269479 [TRT] Tactic: 7733247 Time: 0.158724 [TRT] Tactic: 7798783 Time: 0.20026 [TRT] Tactic: 8191999 Time: 0.299688 [TRT] Tactic: 8257535 Time: 0.220313 [TRT] Tactic: 8323071 Time: 0.191094 [TRT] Tactic: 8650751 Time: 0.282344 [TRT] Tactic: 8716287 Time: 0.152526 [TRT] Tactic: 9109503 Time: 0.243438 [TRT] Tactic: 9568255 Time: 0.210078 [TRT] Tactic: 9895935 Time: 0.201589 [TRT] Tactic: 10223615 Time: 0.222943 [TRT] Tactic: 10354687 Time: 0.243645 [TRT] Tactic: 10551295 Time: 0.18 [TRT] Tactic: 10747903 Time: 0.158073 [TRT] Tactic: 10944511 Time: 0.307552 [TRT] Fastest Tactic: 6225919 Time: 0.146979 [TRT] --------------- Timing Runner: inception_4d/pool_proj + inception_4d/relu_pool_proj (CudnnConvolution) [TRT] Tactic: 0 Time: 0.234688 [TRT] Tactic: 1 Time: 0.198073 [TRT] Tactic: 2 Time: 0.505938 [TRT] Tactic: 4 skipped. Scratch requested: 76808192, available: 33554432 [TRT] Tactic: 5 Time: 2.07578 [TRT] Fastest Tactic: 1 Time: 0.198073 [TRT] --------------- Timing Runner: inception_4d/pool_proj + inception_4d/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4d/pool_proj + inception_4d/relu_pool_proj (CaskConvolution) [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.144375 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v0 Tactic: 1698681053543049347 [TRT] Tactic: 1698681053543049347 Time: 0.130078 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.183307 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.113828 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.183073 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v1 Tactic: 5326823351883942011 [TRT] Tactic: 5326823351883942011 Time: 0.176354 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v0 Tactic: 5500448035057547314 [TRT] Tactic: 5500448035057547314 Time: 0.140859 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.116458 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.137213 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.185182 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v1 Tactic: -6576203419454146580 [TRT] Tactic: -6576203419454146580 Time: 0.127838 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.142812 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.135339 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.151042 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.145885 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.17974 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v0 Tactic: -377491875521947884 [TRT] Tactic: -377491875521947884 Time: 0.181146 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v1 Tactic: -37215280111360163 [TRT] Tactic: -37215280111360163 Time: 0.112787 [TRT] Fastest Tactic: -37215280111360163 Time: 0.112787 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -37215280111360163 [TRT] *************** Autotuning format combination: Float(100352,1,7168,512) -> Float(103488,1,7392,528) *************** [TRT] --------------- Timing Runner: inception_4d/pool_proj + inception_4d/relu_pool_proj (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4d/pool_proj + inception_4d/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4d/pool_proj + inception_4d/relu_pool_proj (CaskConvolution) [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 3886731678879822788 [TRT] Tactic: 3886731678879822788 Time: 0.0970055 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 6629944304117643200 [TRT] Tactic: 6629944304117643200 Time: 0.152579 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.154792 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.0982545 [TRT] Fastest Tactic: 3886731678879822788 Time: 0.0970055 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 3886731678879822788 [TRT] *************** Autotuning format combination: Half(100352,196,14,1) -> Half(103488,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/pool_proj + inception_4d/relu_pool_proj (CudnnConvolution) [TRT] Tactic: 0 Time: 0.240157 [TRT] Tactic: 1 Time: 0.237265 [TRT] Tactic: 2 Time: 0.501822 [TRT] Tactic: 4 skipped. Scratch requested: 76808192, available: 33554432 [TRT] Tactic: 5 Time: 1.95346 [TRT] Fastest Tactic: 1 Time: 0.237265 [TRT] --------------- Timing Runner: inception_4d/pool_proj + inception_4d/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4d/pool_proj + inception_4d/relu_pool_proj (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 1 [TRT] *************** Autotuning format combination: Half(50176,196:2,14,1) -> Half(103488,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/pool_proj + inception_4d/relu_pool_proj (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] *************** Autotuning format combination: Half(50176,196:2,14,1) -> Half(51744,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/pool_proj + inception_4d/relu_pool_proj (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.112682 [TRT] Tactic: 655359 Time: 0.192006 [TRT] Tactic: 786431 Time: 0.140703 [TRT] Tactic: 851967 Time: 0.174922 [TRT] Tactic: 1179647 Time: 0.0879165 [TRT] Tactic: 1310719 Time: 0.245312 [TRT] Tactic: 1376255 Time: 0.114557 [TRT] Tactic: 1441791 Time: 0.121562 [TRT] Tactic: 1507327 Time: 0.219063 [TRT] Tactic: 1638399 Time: 0.145182 [TRT] Tactic: 1835007 Time: 0.137318 [TRT] Tactic: 1900543 Time: 0.204661 [TRT] Tactic: 2097151 Time: 0.142188 [TRT] Tactic: 2162687 Time: 0.116848 [TRT] Tactic: 2293759 Time: 0.117344 [TRT] Tactic: 2359295 Time: 0.11987 [TRT] Tactic: 2686975 Time: 0.207812 [TRT] Tactic: 3080191 Time: 0.150833 [TRT] Tactic: 3342335 Time: 0.173489 [TRT] Tactic: 3407871 Time: 0.100938 [TRT] Tactic: 3538943 Time: 0.0876825 [TRT] Tactic: 3670015 Time: 0.210807 [TRT] Tactic: 3932159 Time: 0.163438 [TRT] Tactic: 3997695 Time: 0.128359 [TRT] Tactic: 4063231 Time: 0.153438 [TRT] Tactic: 4194303 Time: 0.104896 [TRT] Tactic: 4259839 Time: 0.136198 [TRT] Tactic: 4325375 Time: 0.128256 [TRT] Tactic: 4521983 Time: 0.132318 [TRT] Tactic: 4587519 Time: 0.129349 [TRT] Tactic: 4653055 Time: 0.107291 [TRT] Tactic: 4915199 Time: 0.10763 [TRT] Tactic: 4980735 Time: 0.139453 [TRT] Tactic: 5177343 Time: 0.097552 [TRT] Tactic: 5242879 Time: 0.079635 [TRT] Tactic: 5373951 Time: 0.094687 [TRT] Tactic: 5439487 Time: 0.103568 [TRT] Tactic: 5570559 Time: 0.168646 [TRT] Tactic: 5636095 Time: 0.154687 [TRT] Tactic: 5701631 Time: 0.111172 [TRT] Tactic: 5767167 Time: 0.139297 [TRT] Tactic: 5832703 Time: 0.0946095 [TRT] Tactic: 5898239 Time: 0.0896875 [TRT] Tactic: 6029311 Time: 0.108098 [TRT] Tactic: 6225919 Time: 0.075807 [TRT] Tactic: 6291455 Time: 0.087422 [TRT] Tactic: 6422527 Time: 0.141875 [TRT] Tactic: 6750207 Time: 0.104531 [TRT] Tactic: 6815743 Time: 0.0786195 [TRT] Tactic: 6946815 Time: 0.129557 [TRT] Tactic: 7012351 Time: 0.14125 [TRT] Tactic: 7077887 Time: 0.083203 [TRT] Tactic: 7143423 Time: 0.149375 [TRT] Tactic: 7208959 Time: 0.099245 [TRT] Tactic: 7340031 Time: 0.092396 [TRT] Tactic: 7405567 Time: 0.088021 [TRT] Tactic: 7536639 Time: 0.106042 [TRT] Tactic: 7602175 Time: 0.118933 [TRT] Tactic: 7733247 Time: 0.0891925 [TRT] Tactic: 7798783 Time: 0.140105 [TRT] Tactic: 8191999 Time: 0.153412 [TRT] Tactic: 8257535 Time: 0.105912 [TRT] Tactic: 8323071 Time: 0.102474 [TRT] Tactic: 8650751 Time: 0.129297 [TRT] Tactic: 8716287 Time: 0.078646 [TRT] Tactic: 9109503 Time: 0.139688 [TRT] Tactic: 9568255 Time: 0.10763 [TRT] Tactic: 9895935 Time: 0.105911 [TRT] Tactic: 10223615 Time: 0.206719 [TRT] Tactic: 10354687 Time: 0.139558 [TRT] Tactic: 10551295 Time: 0.0953385 [TRT] Tactic: 10747903 Time: 0.0834375 [TRT] Tactic: 10944511 Time: 0.138333 [TRT] Fastest Tactic: 6225919 Time: 0.075807 [TRT] --------------- Timing Runner: inception_4d/pool_proj + inception_4d/relu_pool_proj (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4d/pool_proj + inception_4d/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4d/pool_proj + inception_4d/relu_pool_proj (CaskConvolution) [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_interior_nn_v1 Tactic: 3066127711859985668 [TRT] Tactic: 3066127711859985668 Time: 0.0665885 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.0741405 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.0718485 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.0621875 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] Tactic: 8163473458334948789 Time: 0.0604955 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.095729 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.099297 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.0638545 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_interior_nn_v1 Tactic: -1716393687483585322 [TRT] Tactic: -1716393687483585322 Time: 0.0948175 [TRT] Fastest Tactic: 8163473458334948789 Time: 0.0604955 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 8163473458334948789 [TRT] *************** Autotuning Reformat:Float(103488,196,14,1) -> Float(103488,1,7392,528) *************** [TRT] *************** Autotuning Reformat:Float(103488,196,14,1) -> Half(103488,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(103488,196,14,1) -> Half(51744,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Float(103488,1,7392,528) -> Float(103488,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(103488,1,7392,528) -> Half(103488,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(103488,1,7392,528) -> Half(51744,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(103488,196,14,1) -> Float(103488,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(103488,196,14,1) -> Float(103488,1,7392,528) *************** [TRT] *************** Autotuning Reformat:Half(103488,196,14,1) -> Half(51744,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(51744,196:2,14,1) -> Float(103488,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(51744,196:2,14,1) -> Float(103488,1,7392,528) *************** [TRT] *************** Autotuning Reformat:Half(51744,196:2,14,1) -> Half(103488,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(56448,196,14,1) -> Float(103488,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 1.64513 [TRT] Tactic: 0 Time: 0.008047 [TRT] Fastest Tactic: 0 Time: 0.008047 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(56448,196,14,1) -> Float(103488,1,7392,528) *************** [TRT] --------------- Timing Runner: inception_4d/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0364585 [TRT] Tactic: 0 Time: 0.0374475 [TRT] Fastest Tactic: 1002 Time: 0.0364585 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(56448,196,14,1) -> Half(103488,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.717917 [TRT] Tactic: 0 Time: 0.033515 [TRT] Fastest Tactic: 0 Time: 0.033515 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(56448,196,14,1) -> Half(51744,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0417185 [TRT] Tactic: 0 Time: 0.0403385 [TRT] Fastest Tactic: 0 Time: 0.0403385 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(56448,1,4032,288) -> Float(103488,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0451045 [TRT] Tactic: 0 Time: 0.0336195 [TRT] Fastest Tactic: 0 Time: 0.0336195 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(56448,1,4032,288) -> Float(103488,1,7392,528) *************** [TRT] --------------- Timing Runner: inception_4d/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.065156 [TRT] Tactic: 0 Time: 0.036146 [TRT] Fastest Tactic: 0 Time: 0.036146 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(56448,1,4032,288) -> Half(103488,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0334895 [TRT] Tactic: 0 Time: 0.0347135 [TRT] Fastest Tactic: 1002 Time: 0.0334895 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(56448,1,4032,288) -> Half(51744,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.035807 [TRT] Tactic: 0 Time: 0.0403905 [TRT] Fastest Tactic: 1002 Time: 0.035807 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(56448,196,14,1) -> Float(103488,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.735286 [TRT] Tactic: 0 Time: 0.0346615 [TRT] Fastest Tactic: 0 Time: 0.0346615 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(56448,196,14,1) -> Float(103488,1,7392,528) *************** [TRT] --------------- Timing Runner: inception_4d/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0311455 [TRT] Tactic: 0 Time: 0.038047 [TRT] Fastest Tactic: 1002 Time: 0.0311455 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(56448,196,14,1) -> Half(103488,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.732031 [TRT] Tactic: 0 Time: 0.005677 [TRT] Fastest Tactic: 0 Time: 0.005677 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(56448,196,14,1) -> Half(51744,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.030807 [TRT] Tactic: 0 Time: 0.0224745 [TRT] Fastest Tactic: 0 Time: 0.0224745 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(28224,196:2,14,1) -> Float(103488,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.041094 [TRT] Tactic: 0 Time: 0.0396355 [TRT] Fastest Tactic: 0 Time: 0.0396355 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(28224,196:2,14,1) -> Float(103488,1,7392,528) *************** [TRT] --------------- Timing Runner: inception_4d/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0311455 [TRT] Tactic: 0 Time: 0.0426305 [TRT] Fastest Tactic: 1002 Time: 0.0311455 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(28224,196:2,14,1) -> Half(103488,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0519275 [TRT] Tactic: 0 Time: 0.035807 [TRT] Fastest Tactic: 0 Time: 0.035807 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(28224,196:2,14,1) -> Half(51744,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4d/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0403385 [TRT] Tactic: 0 Time: 0.0055465 [TRT] Fastest Tactic: 0 Time: 0.0055465 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(103488,196,14,1) -> Float(103488,1,7392,528) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.154792 [TRT] Tactic: 0 Time: 0.160105 [TRT] Fastest Tactic: 1002 Time: 0.154792 [TRT] *************** Autotuning Reformat:Float(103488,196,14,1) -> Half(103488,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 2.78297 [TRT] Tactic: 0 Time: 0.113047 [TRT] Fastest Tactic: 0 Time: 0.113047 [TRT] *************** Autotuning Reformat:Float(103488,196,14,1) -> Half(51744,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.170886 [TRT] Tactic: 0 Time: 0.089635 [TRT] Fastest Tactic: 0 Time: 0.089635 [TRT] *************** Autotuning Reformat:Float(103488,1,7392,528) -> Float(103488,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.184141 [TRT] Tactic: 0 Time: 0.148151 [TRT] Fastest Tactic: 0 Time: 0.148151 [TRT] *************** Autotuning Reformat:Float(103488,1,7392,528) -> Half(103488,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.122422 [TRT] Tactic: 0 Time: 0.148464 [TRT] Fastest Tactic: 1002 Time: 0.122422 [TRT] *************** Autotuning Reformat:Float(103488,1,7392,528) -> Half(51744,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.149584 [TRT] Tactic: 0 Time: 0.175938 [TRT] Fastest Tactic: 1002 Time: 0.149584 [TRT] *************** Autotuning Reformat:Half(103488,196,14,1) -> Float(103488,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 2.86883 [TRT] Tactic: 0 Time: 0.0949485 [TRT] Fastest Tactic: 0 Time: 0.0949485 [TRT] *************** Autotuning Reformat:Half(103488,196,14,1) -> Float(103488,1,7392,528) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.113907 [TRT] Tactic: 0 Time: 0.159999 [TRT] Fastest Tactic: 1002 Time: 0.113907 [TRT] *************** Autotuning Reformat:Half(103488,196,14,1) -> Half(51744,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.124453 [TRT] Tactic: 0 Time: 0.08862 [TRT] Fastest Tactic: 0 Time: 0.08862 [TRT] *************** Autotuning Reformat:Half(51744,196:2,14,1) -> Float(103488,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.156823 [TRT] Tactic: 0 Time: 0.077656 [TRT] Fastest Tactic: 0 Time: 0.077656 [TRT] *************** Autotuning Reformat:Half(51744,196:2,14,1) -> Float(103488,1,7392,528) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.114921 [TRT] Tactic: 0 Time: 0.186355 [TRT] Fastest Tactic: 1002 Time: 0.114921 [TRT] *************** Autotuning Reformat:Half(51744,196:2,14,1) -> Half(103488,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.232343 [TRT] Tactic: 0 Time: 0.0752865 [TRT] Fastest Tactic: 0 Time: 0.0752865 [TRT] *************** Autotuning format combination: Float(103488,196,14,1) -> Float(87808,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 1.27846 [TRT] Tactic: 655359 Time: 1.14135 [TRT] Tactic: 786431 Time: 1.41195 [TRT] Tactic: 851967 Time: 1.7394 [TRT] Tactic: 1179647 Time: 1.61102 [TRT] Tactic: 1310719 Time: 2.8737 [TRT] Tactic: 1376255 Time: 1.11044 [TRT] Tactic: 1441791 Time: 1.86135 [TRT] Tactic: 1507327 Time: 1.828 [TRT] Tactic: 1638399 Time: 2.10086 [TRT] Tactic: 1835007 Time: 1.46969 [TRT] Tactic: 1900543 Time: 1.51854 [TRT] Tactic: 2162687 Time: 1.1294 [TRT] Tactic: 2293759 Time: 1.22336 [TRT] Tactic: 2359295 Time: 1.40133 [TRT] Tactic: 2686975 Time: 1.377 [TRT] Tactic: 3080191 Time: 1.37763 [TRT] Tactic: 3342335 Time: 1.63805 [TRT] Tactic: 3407871 Time: 1.28354 [TRT] Tactic: 3538943 Time: 1.37388 [TRT] Tactic: 3670015 Time: 0.926484 [TRT] Tactic: 3932159 Time: 2.05237 [TRT] Tactic: 3997695 Time: 1.43956 [TRT] Tactic: 4063231 Time: 1.52201 [TRT] Tactic: 4194303 Time: 1.37125 [TRT] Tactic: 4325375 Time: 1.73104 [TRT] Tactic: 4521983 Time: 1.68779 [TRT] Tactic: 4587519 Time: 1.57919 [TRT] Tactic: 4653055 Time: 1.55753 [TRT] Tactic: 4915199 Time: 1.33393 [TRT] Tactic: 4980735 Time: 1.74125 [TRT] Tactic: 5177343 Time: 1.94576 [TRT] Tactic: 5242879 Time: 1.18492 [TRT] Tactic: 5373951 Time: 1.86125 [TRT] Tactic: 5439487 Time: 1.50198 [TRT] Tactic: 5570559 Time: 1.18185 [TRT] Tactic: 5636095 Time: 1.51102 [TRT] Tactic: 5701631 Time: 1.44328 [TRT] Tactic: 5767167 Time: 2.39112 [TRT] Tactic: 5832703 Time: 1.30159 [TRT] Tactic: 5898239 Time: 1.13771 [TRT] Tactic: 6029311 Time: 1.17672 [TRT] Tactic: 6225919 Time: 1.27393 [TRT] Tactic: 6291455 Time: 1.61357 [TRT] Tactic: 6422527 Time: 1.41359 [TRT] Tactic: 6750207 Time: 1.38016 [TRT] Tactic: 6815743 Time: 1.2931 [TRT] Tactic: 6946815 Time: 1.91417 [TRT] Tactic: 7077887 Time: 1.3757 [TRT] Tactic: 7143423 Time: 2.16643 [TRT] Tactic: 7208959 Time: 1.52651 [TRT] Tactic: 7340031 Time: 1.24 [TRT] Tactic: 7405567 Time: 1.40815 [TRT] Tactic: 7536639 Time: 1.60391 [TRT] Tactic: 7602175 Time: 1.79688 [TRT] Tactic: 7733247 Time: 1.27779 [TRT] Tactic: 7798783 Time: 1.40961 [TRT] Tactic: 8191999 Time: 2.24758 [TRT] Tactic: 8257535 Time: 1.39682 [TRT] Tactic: 8323071 Time: 1.32594 [TRT] Tactic: 8650751 Time: 2.06997 [TRT] Tactic: 8716287 Time: 1.45125 [TRT] Tactic: 9568255 Time: 1.33083 [TRT] Tactic: 9895935 Time: 1.3513 [TRT] Tactic: 10223615 Time: 1.37396 [TRT] Tactic: 10354687 Time: 1.71698 [TRT] Tactic: 10551295 Time: 1.28708 [TRT] Tactic: 10747903 Time: 1.16685 [TRT] Tactic: 10944511 Time: 1.74534 [TRT] Fastest Tactic: 3670015 Time: 0.926484 [TRT] --------------- Timing Runner: inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce (CudnnConvolution) [TRT] Tactic: 0 Time: 1.35003 [TRT] Tactic: 1 Time: 1.13552 [TRT] Tactic: 2 Time: 1.38773 [TRT] Tactic: 4 skipped. Scratch requested: 547160064, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 34028544, available: 33554432 [TRT] Fastest Tactic: 1 Time: 1.13552 [TRT] --------------- Timing Runner: inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce (CaskConvolution) [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.848151 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v0 Tactic: 1698681053543049347 [TRT] Tactic: 1698681053543049347 Time: 0.776224 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.740026 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.64677 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.731614 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v1 Tactic: 5326823351883942011 [TRT] Tactic: 5326823351883942011 Time: 0.707004 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v0 Tactic: 5500448035057547314 [TRT] Tactic: 5500448035057547314 Time: 0.719218 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.675182 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.865911 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.742527 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v1 Tactic: -6576203419454146580 [TRT] Tactic: -6576203419454146580 Time: 0.758932 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.908646 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.797969 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.752709 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.741432 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.721146 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v0 Tactic: -377491875521947884 [TRT] Tactic: -377491875521947884 Time: 0.724219 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v1 Tactic: -37215280111360163 [TRT] Tactic: -37215280111360163 Time: 0.636328 [TRT] Fastest Tactic: -37215280111360163 Time: 0.636328 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -37215280111360163 [TRT] *************** Autotuning format combination: Float(103488,1,7392,528) -> Float(87808,1,6272,448) *************** [TRT] --------------- Timing Runner: inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce (CaskConvolution) [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 3886731678879822788 [TRT] Tactic: 3886731678879822788 Time: 0.664427 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 6629944304117643200 [TRT] Tactic: 6629944304117643200 Time: 1.11948 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 1.12177 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.650885 [TRT] Fastest Tactic: -7394439838318485025 Time: 0.650885 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(103488,196,14,1) -> Half(87808,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce (CudnnConvolution) [TRT] Tactic: 0 Time: 1.35513 [TRT] Tactic: 1 Time: 1.18424 [TRT] Tactic: 2 Time: 1.35977 [TRT] Tactic: 4 skipped. Scratch requested: 547160064, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 34028544, available: 33554432 [TRT] Fastest Tactic: 1 Time: 1.18424 [TRT] --------------- Timing Runner: inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] Setting workspace to 34028544enables more tactics for profiling [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 1 [TRT] *************** Autotuning format combination: Half(51744,196:2,14,1) -> Half(87808,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] *************** Autotuning format combination: Half(51744,196:2,14,1) -> Half(43904,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.662109 [TRT] Tactic: 655359 Time: 0.67849 [TRT] Tactic: 786431 Time: 0.884166 [TRT] Tactic: 851967 Time: 0.935781 [TRT] Tactic: 1310719 Time: 1.42831 [TRT] Tactic: 1376255 Time: 0.547214 [TRT] Tactic: 1507327 Time: 0.946249 [TRT] Tactic: 1638399 Time: 1.10333 [TRT] Tactic: 1835007 Time: 0.927604 [TRT] Tactic: 1900543 Time: 0.810703 [TRT] Tactic: 2162687 Time: 0.574974 [TRT] Tactic: 2293759 Time: 0.652969 [TRT] Tactic: 2359295 Time: 0.716875 [TRT] Tactic: 2686975 Time: 1.1343 [TRT] Tactic: 3080191 Time: 0.724818 [TRT] Tactic: 3342335 Time: 0.870052 [TRT] Tactic: 3407871 Time: 0.638333 [TRT] Tactic: 3538943 Time: 0.66513 [TRT] Tactic: 3670015 Time: 0.725 [TRT] Tactic: 3932159 Time: 0.895677 [TRT] Tactic: 4063231 Time: 0.798698 [TRT] Tactic: 4194303 Time: 0.77388 [TRT] Tactic: 4325375 Time: 0.885781 [TRT] Tactic: 4521983 Time: 0.856745 [TRT] Tactic: 4980735 Time: 0.892136 [TRT] Tactic: 5242879 Time: 0.57224 [TRT] Tactic: 5439487 Time: 0.844427 [TRT] Tactic: 5570559 Time: 0.771484 [TRT] Tactic: 5636095 Time: 0.798098 [TRT] Tactic: 5701631 Time: 0.682343 [TRT] Tactic: 5767167 Time: 1.09875 [TRT] Tactic: 5832703 Time: 0.623073 [TRT] Tactic: 6029311 Time: 0.603125 [TRT] Tactic: 6225919 Time: 0.587734 [TRT] Tactic: 6422527 Time: 0.696146 [TRT] Tactic: 6815743 Time: 0.635208 [TRT] Tactic: 6946815 Time: 0.933177 [TRT] Tactic: 7077887 Time: 0.634453 [TRT] Tactic: 7143423 Time: 1.06977 [TRT] Tactic: 7208959 Time: 0.74711 [TRT] Tactic: 7405567 Time: 0.675521 [TRT] Tactic: 7536639 Time: 0.794479 [TRT] Tactic: 7602175 Time: 0.843646 [TRT] Tactic: 7733247 Time: 0.674479 [TRT] Tactic: 7798783 Time: 0.877552 [TRT] Tactic: 8191999 Time: 1.11753 [TRT] Tactic: 8323071 Time: 0.761068 [TRT] Tactic: 8650751 Time: 0.988594 [TRT] Tactic: 8716287 Time: 0.684348 [TRT] Tactic: 9895935 Time: 0.771563 [TRT] Tactic: 10223615 Time: 1.13057 [TRT] Tactic: 10551295 Time: 0.642084 [TRT] Tactic: 10747903 Time: 0.586771 [TRT] Tactic: 10944511 Time: 0.890964 [TRT] Fastest Tactic: 1376255 Time: 0.547214 [TRT] --------------- Timing Runner: inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce (CaskConvolution) [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_interior_nn_v1 Tactic: 3066127711859985668 [TRT] Tactic: 3066127711859985668 Time: 0.383438 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.423776 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.402968 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.344609 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] Tactic: 8163473458334948789 Time: 0.326589 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.372292 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.382735 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.337786 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_interior_nn_v1 Tactic: -1716393687483585322 [TRT] Tactic: -1716393687483585322 Time: 0.367682 [TRT] Fastest Tactic: 8163473458334948789 Time: 0.326589 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 8163473458334948789 [TRT] *************** Autotuning Reformat:Float(87808,196,14,1) -> Float(87808,1,6272,448) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.120704 [TRT] Tactic: 0 Time: 0.136224 [TRT] Fastest Tactic: 1002 Time: 0.120704 [TRT] *************** Autotuning Reformat:Float(87808,196,14,1) -> Half(87808,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 2.36333 [TRT] Tactic: 0 Time: 0.09625 [TRT] Fastest Tactic: 0 Time: 0.09625 [TRT] *************** Autotuning Reformat:Float(87808,196,14,1) -> Half(43904,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.145052 [TRT] Tactic: 0 Time: 0.076536 [TRT] Fastest Tactic: 0 Time: 0.076536 [TRT] *************** Autotuning Reformat:Float(87808,1,6272,448) -> Float(87808,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.153698 [TRT] Tactic: 0 Time: 0.131198 [TRT] Fastest Tactic: 0 Time: 0.131198 [TRT] *************** Autotuning Reformat:Float(87808,1,6272,448) -> Half(87808,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.103255 [TRT] Tactic: 0 Time: 0.12862 [TRT] Fastest Tactic: 1002 Time: 0.103255 [TRT] *************** Autotuning Reformat:Float(87808,1,6272,448) -> Half(43904,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.126198 [TRT] Tactic: 0 Time: 0.150026 [TRT] Fastest Tactic: 1002 Time: 0.126198 [TRT] *************** Autotuning Reformat:Half(87808,196,14,1) -> Float(87808,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 2.43563 [TRT] Tactic: 0 Time: 0.0809635 [TRT] Fastest Tactic: 0 Time: 0.0809635 [TRT] *************** Autotuning Reformat:Half(87808,196,14,1) -> Float(87808,1,6272,448) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.095417 [TRT] Tactic: 0 Time: 0.136901 [TRT] Fastest Tactic: 1002 Time: 0.095417 [TRT] *************** Autotuning Reformat:Half(87808,196,14,1) -> Half(43904,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.106536 [TRT] Tactic: 0 Time: 0.076406 [TRT] Fastest Tactic: 0 Time: 0.076406 [TRT] *************** Autotuning Reformat:Half(43904,196:2,14,1) -> Float(87808,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.135105 [TRT] Tactic: 0 Time: 0.066406 [TRT] Fastest Tactic: 0 Time: 0.066406 [TRT] *************** Autotuning Reformat:Half(43904,196:2,14,1) -> Float(87808,1,6272,448) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.095911 [TRT] Tactic: 0 Time: 0.158985 [TRT] Fastest Tactic: 1002 Time: 0.095911 [TRT] *************** Autotuning Reformat:Half(43904,196:2,14,1) -> Half(87808,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.193724 [TRT] Tactic: 0 Time: 0.0651565 [TRT] Fastest Tactic: 0 Time: 0.0651565 [TRT] *************** Autotuning Reformat:Float(87808,196,14,1) -> Float(87808,1,6272,448) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.05362 [TRT] Tactic: 0 Time: 0.051432 [TRT] Fastest Tactic: 0 Time: 0.051432 [TRT] *************** Autotuning Reformat:Float(87808,196,14,1) -> Half(87808,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.02242 [TRT] Tactic: 0 Time: 0.047395 [TRT] Fastest Tactic: 0 Time: 0.047395 [TRT] *************** Autotuning Reformat:Float(87808,196,14,1) -> Half(43904,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.056328 [TRT] Tactic: 0 Time: 0.0563025 [TRT] Fastest Tactic: 0 Time: 0.0563025 [TRT] *************** Autotuning Reformat:Float(87808,1,6272,448) -> Float(87808,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0641405 [TRT] Tactic: 0 Time: 0.047969 [TRT] Fastest Tactic: 0 Time: 0.047969 [TRT] *************** Autotuning Reformat:Float(87808,1,6272,448) -> Half(87808,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0409375 [TRT] Tactic: 0 Time: 0.047604 [TRT] Fastest Tactic: 1002 Time: 0.0409375 [TRT] *************** Autotuning Reformat:Float(87808,1,6272,448) -> Half(43904,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0498175 [TRT] Tactic: 0 Time: 0.057031 [TRT] Fastest Tactic: 1002 Time: 0.0498175 [TRT] *************** Autotuning Reformat:Half(87808,196,14,1) -> Float(87808,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.04805 [TRT] Tactic: 0 Time: 0.0312505 [TRT] Fastest Tactic: 0 Time: 0.0312505 [TRT] *************** Autotuning Reformat:Half(87808,196,14,1) -> Float(87808,1,6272,448) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.036641 [TRT] Tactic: 0 Time: 0.0514845 [TRT] Fastest Tactic: 1002 Time: 0.036641 [TRT] *************** Autotuning Reformat:Half(87808,196,14,1) -> Half(43904,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.040833 [TRT] Tactic: 0 Time: 0.028646 [TRT] Fastest Tactic: 0 Time: 0.028646 [TRT] *************** Autotuning Reformat:Half(43904,196:2,14,1) -> Float(87808,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0526305 [TRT] Tactic: 0 Time: 0.0262765 [TRT] Fastest Tactic: 0 Time: 0.0262765 [TRT] *************** Autotuning Reformat:Half(43904,196:2,14,1) -> Float(87808,1,6272,448) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.037448 [TRT] Tactic: 0 Time: 0.0589325 [TRT] Fastest Tactic: 1002 Time: 0.037448 [TRT] *************** Autotuning Reformat:Half(43904,196:2,14,1) -> Half(87808,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.075208 [TRT] Tactic: 0 Time: 0.049662 [TRT] Fastest Tactic: 0 Time: 0.049662 [TRT] *************** Autotuning format combination: Float(87808,196,14,1) -> Float(163072,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/3x3 + inception_4e/relu_3x3 (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4e/3x3 + inception_4e/relu_3x3 (FusedConvActConvolution) [TRT] Tactic: 524287 Time: 1.25966 [TRT] Tactic: 720895 Time: 1.37867 [TRT] Tactic: 983039 Time: 1.17362 [TRT] Tactic: 1048575 Time: 1.28662 [TRT] Tactic: 1703935 Time: 1.26646 [TRT] Tactic: 1769471 Time: 1.19302 [TRT] Tactic: 1966079 Time: 1.2726 [TRT] Tactic: 2031615 Time: 1.24427 [TRT] Tactic: 2228223 Time: 1.62167 [TRT] Tactic: 2424831 Time: 1.7918 [TRT] Tactic: 2621439 Time: 1.52635 [TRT] Tactic: 2752511 Time: 1.13294 [TRT] Tactic: 2818047 Time: 1.86336 [TRT] Tactic: 2883583 Time: 1.46289 [TRT] Tactic: 3014655 Time: 1.17922 [TRT] Tactic: 3145727 Time: 1.10268 [TRT] Tactic: 3473407 Time: 1.63326 [TRT] Tactic: 3604479 Time: 1.16159 [TRT] Tactic: 3735551 Time: 1.88812 [TRT] Tactic: 4390911 Time: 1.13253 [TRT] Tactic: 5046271 Time: 1.22984 [TRT] Tactic: 5963775 Time: 1.15823 [TRT] Tactic: 6160383 Time: 1.23156 [TRT] Tactic: 6488063 Time: 1.09784 [TRT] Tactic: 6881279 Time: 1.17276 [TRT] Tactic: 7274495 Time: 1.53859 [TRT] Tactic: 7864319 Time: 1.59359 [TRT] Tactic: 7995391 Time: 1.35922 [TRT] Tactic: 8585215 Time: 1.13443 [TRT] Tactic: 8847359 Time: 1.18581 [TRT] Tactic: 8978431 Time: 1.18612 [TRT] Tactic: 9043967 Time: 1.08 [TRT] Tactic: 9175039 Time: 1.16172 [TRT] Tactic: 9502719 Time: 1.18159 [TRT] Tactic: 9830399 Time: 1.40851 [TRT] Tactic: 9961471 Time: 1.32057 [TRT] Tactic: 10027007 Time: 1.10544 [TRT] Tactic: 10092543 Time: 1.1269 [TRT] Tactic: 10289151 Time: 1.27706 [TRT] Tactic: 10485759 Time: 1.11578 [TRT] Tactic: 10682367 Time: 1.53732 [TRT] Tactic: 10813439 Time: 1.33773 [TRT] Fastest Tactic: 9043967 Time: 1.08 [TRT] --------------- Timing Runner: inception_4e/3x3 + inception_4e/relu_3x3 (CudnnConvolution) [TRT] Tactic: 0 Time: 2.57633 [TRT] Tactic: 1 Time: 1.42669 [TRT] Tactic: 2 Time: 2.27651 [TRT] Tactic: 4 skipped. Scratch requested: 120176640, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 224911360, available: 33554432 [TRT] Tactic: 6 Time: 1.09195 [TRT] Fastest Tactic: 6 Time: 1.09195 [TRT] --------------- Timing Runner: inception_4e/3x3 + inception_4e/relu_3x3 (CaskConvolution) [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 1.67115 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v0 Tactic: 1754984623894446479 [TRT] Tactic: 1754984623894446479 Time: 1.77852 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v0 Tactic: 3611739942397549984 [TRT] Tactic: 3611739942397549984 Time: 1.45844 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148n_nt_v1 Tactic: 3827454225649558724 [TRT] Tactic: 3827454225649558724 Time: 0.974948 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v1 Tactic: 4337000649858996379 [TRT] Tactic: 4337000649858996379 Time: 1.29625 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 1.4662 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 1.20568 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 1.4245 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 5921334924264294896 [TRT] Tactic: 5921334924264294896 Time: 0.74987 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 1.27581 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 1.52609 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148t_nt_v1 Tactic: 7852627285308570038 [TRT] Tactic: 7852627285308570038 Time: 0.935521 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v1 Tactic: -9137461792520977713 [TRT] Tactic: -9137461792520977713 Time: 1.4712 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148n_nt_v0 Tactic: -8776506421218919509 [TRT] Tactic: -8776506421218919509 Time: 0.915235 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 1.46677 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v0 Tactic: -8133971918129952780 [TRT] Tactic: -8133971918129952780 Time: 1.43969 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v1 Tactic: -6092040395344634144 [TRT] Tactic: -6092040395344634144 Time: 1.74406 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 1.76992 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 1.41375 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148t_nt_v0 Tactic: -2318106587342035239 [TRT] Tactic: -2318106587342035239 Time: 0.900182 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_mobile_relu_tile148t_nt_v0 Tactic: -1343271414618805657 [TRT] Tactic: -1343271414618805657 Time: 0.677344 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 1.43214 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 1.36789 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 1.40297 [TRT] Fastest Tactic: -1343271414618805657 Time: 0.677344 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -1343271414618805657 [TRT] *************** Autotuning format combination: Float(87808,1,6272,448) -> Float(163072,1,11648,832) *************** [TRT] --------------- Timing Runner: inception_4e/3x3 + inception_4e/relu_3x3 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4e/3x3 + inception_4e/relu_3x3 (CaskConvolution) [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 1.47221 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 1.17141 [TRT] Fastest Tactic: -7394439838318485025 Time: 1.17141 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(87808,196,14,1) -> Half(163072,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/3x3 + inception_4e/relu_3x3 (CudnnConvolution) [TRT] Tactic: 0 Time: 2.75432 [TRT] Tactic: 1 Time: 2.85643 [TRT] Tactic: 2 Time: 2.22417 [TRT] Tactic: 4 skipped. Scratch requested: 120176640, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 224911360, available: 33554432 [TRT] Tactic: 6 Time: 2.75135 [TRT] Fastest Tactic: 2 Time: 2.22417 [TRT] --------------- Timing Runner: inception_4e/3x3 + inception_4e/relu_3x3 (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 2 [TRT] *************** Autotuning format combination: Half(43904,196:2,14,1) -> Half(81536,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/3x3 + inception_4e/relu_3x3 (FusedConvActConvolution) [TRT] Tactic: 524287 Time: 0.660286 [TRT] Tactic: 720895 Time: 0.811119 [TRT] Tactic: 983039 Time: 0.64651 [TRT] Tactic: 1048575 Time: 0.708463 [TRT] Tactic: 1703935 Time: 0.718333 [TRT] Tactic: 1769471 Time: 4.17799 [TRT] Tactic: 1966079 Time: 0.795833 [TRT] Tactic: 2031615 Time: 0.666823 [TRT] Tactic: 2228223 Time: 0.895339 [TRT] Tactic: 2424831 Time: 1.33229 [TRT] Tactic: 2621439 Time: 0.820678 [TRT] Tactic: 2752511 Time: 0.630078 [TRT] Tactic: 2818047 Time: 1.01508 [TRT] Tactic: 2883583 Time: 0.916484 [TRT] Tactic: 3014655 Time: 0.670416 [TRT] Tactic: 3145727 Time: 0.637448 [TRT] Tactic: 3473407 Time: 0.964609 [TRT] Tactic: 3604479 Time: 0.660782 [TRT] Tactic: 3735551 Time: 0.924688 [TRT] Tactic: 4390911 Time: 0.680547 [TRT] Tactic: 5046271 Time: 0.663646 [TRT] Tactic: 5963775 Time: 0.635365 [TRT] Tactic: 6160383 Time: 0.706172 [TRT] Tactic: 6488063 Time: 0.603047 [TRT] Tactic: 6881279 Time: 0.61987 [TRT] Tactic: 7274495 Time: 0.865703 [TRT] Tactic: 7864319 Time: 0.877005 [TRT] Tactic: 7995391 Time: 0.777604 [TRT] Tactic: 8585215 Time: 0.634948 [TRT] Tactic: 8847359 Time: 0.64638 [TRT] Tactic: 8978431 Time: 0.685521 [TRT] Tactic: 9043967 Time: 0.587214 [TRT] Tactic: 9175039 Time: 0.660781 [TRT] Tactic: 9502719 Time: 0.669688 [TRT] Tactic: 9830399 Time: 0.751276 [TRT] Tactic: 9961471 Time: 0.726928 [TRT] Tactic: 10027007 Time: 0.609036 [TRT] Tactic: 10092543 Time: 0.678151 [TRT] Tactic: 10289151 Time: 0.798568 [TRT] Tactic: 10485759 Time: 0.604375 [TRT] Tactic: 10682367 Time: 0.804531 [TRT] Tactic: 10813439 Time: 0.806693 [TRT] Fastest Tactic: 9043967 Time: 0.587214 [TRT] --------------- Timing Runner: inception_4e/3x3 + inception_4e/relu_3x3 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4e/3x3 + inception_4e/relu_3x3 (CaskConvolution) [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.83302 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_large_nn_v1 Tactic: 3650389455493082349 [TRT] Tactic: 3650389455493082349 Time: 0.88315 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_winograd_fp16x2_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 4772821744921268633 [TRT] Tactic: 4772821744921268633 Time: 0.395781 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.718932 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.65211 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_large_nn_v1 Tactic: -6490690591794140522 [TRT] Tactic: -6490690591794140522 Time: 0.660651 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_large_nn_v1 Tactic: -4686027666808657977 [TRT] Tactic: -4686027666808657977 Time: 0.750026 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.712005 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.733567 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.611068 [TRT] Fastest Tactic: 4772821744921268633 Time: 0.395781 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 4772821744921268633 [TRT] *************** Autotuning Reformat:Float(163072,196,14,1) -> Float(163072,1,11648,832) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0877345 [TRT] Tactic: 0 Time: 0.098802 [TRT] Fastest Tactic: 1002 Time: 0.0877345 [TRT] *************** Autotuning Reformat:Float(163072,196,14,1) -> Half(163072,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 2.03714 [TRT] Tactic: 0 Time: 0.0914585 [TRT] Fastest Tactic: 0 Time: 0.0914585 [TRT] *************** Autotuning Reformat:Float(163072,196,14,1) -> Half(81536,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.105 [TRT] Tactic: 0 Time: 0.055807 [TRT] Fastest Tactic: 0 Time: 0.055807 [TRT] *************** Autotuning Reformat:Float(163072,1,11648,832) -> Float(163072,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.108073 [TRT] Tactic: 0 Time: 0.093672 [TRT] Fastest Tactic: 0 Time: 0.093672 [TRT] *************** Autotuning Reformat:Float(163072,1,11648,832) -> Half(163072,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.075964 [TRT] Tactic: 0 Time: 0.093203 [TRT] Fastest Tactic: 1002 Time: 0.075964 [TRT] *************** Autotuning Reformat:Float(163072,1,11648,832) -> Half(81536,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0921355 [TRT] Tactic: 0 Time: 0.109062 [TRT] Fastest Tactic: 1002 Time: 0.0921355 [TRT] *************** Autotuning Reformat:Half(163072,196,14,1) -> Float(163072,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 2.08687 [TRT] Tactic: 0 Time: 0.091511 [TRT] Fastest Tactic: 0 Time: 0.091511 [TRT] *************** Autotuning Reformat:Half(163072,196,14,1) -> Float(163072,1,11648,832) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.070104 [TRT] Tactic: 0 Time: 0.0998965 [TRT] Fastest Tactic: 1002 Time: 0.070104 [TRT] *************** Autotuning Reformat:Half(163072,196,14,1) -> Half(81536,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0758075 [TRT] Tactic: 0 Time: 0.0548695 [TRT] Fastest Tactic: 0 Time: 0.0548695 [TRT] *************** Autotuning Reformat:Half(81536,196:2,14,1) -> Float(163072,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.099114 [TRT] Tactic: 0 Time: 0.107213 [TRT] Fastest Tactic: 1002 Time: 0.099114 [TRT] *************** Autotuning Reformat:Half(81536,196:2,14,1) -> Float(163072,1,11648,832) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.069948 [TRT] Tactic: 0 Time: 0.114765 [TRT] Fastest Tactic: 1002 Time: 0.069948 [TRT] *************** Autotuning Reformat:Half(81536,196:2,14,1) -> Half(163072,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.13724 [TRT] Tactic: 0 Time: 0.0968485 [TRT] Fastest Tactic: 0 Time: 0.0968485 [TRT] *************** Autotuning Reformat:Float(87808,196,14,1) -> Float(87808,1,6272,448) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0241665 [TRT] Tactic: 0 Time: 0.012526 [TRT] Fastest Tactic: 0 Time: 0.012526 [TRT] *************** Autotuning Reformat:Float(87808,196,14,1) -> Half(87808,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.209531 [TRT] Tactic: 0 Time: 0.0126565 [TRT] Fastest Tactic: 0 Time: 0.0126565 [TRT] *************** Autotuning Reformat:Float(87808,196,14,1) -> Half(43904,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0150785 [TRT] Tactic: 0 Time: 0.014844 [TRT] Fastest Tactic: 0 Time: 0.014844 [TRT] *************** Autotuning Reformat:Float(87808,1,6272,448) -> Float(87808,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.026666 [TRT] Tactic: 0 Time: 0.012578 [TRT] Fastest Tactic: 0 Time: 0.012578 [TRT] *************** Autotuning Reformat:Float(87808,1,6272,448) -> Half(87808,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.013047 [TRT] Tactic: 0 Time: 0.0125005 [TRT] Fastest Tactic: 0 Time: 0.0125005 [TRT] *************** Autotuning Reformat:Float(87808,1,6272,448) -> Half(43904,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.014896 [TRT] Tactic: 0 Time: 0.0148175 [TRT] Fastest Tactic: 0 Time: 0.0148175 [TRT] *************** Autotuning Reformat:Half(87808,196,14,1) -> Float(87808,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.215365 [TRT] Tactic: 0 Time: 0.010234 [TRT] Fastest Tactic: 0 Time: 0.010234 [TRT] *************** Autotuning Reformat:Half(87808,196,14,1) -> Float(87808,1,6272,448) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.012552 [TRT] Tactic: 0 Time: 0.0127345 [TRT] Fastest Tactic: 1002 Time: 0.012552 [TRT] *************** Autotuning Reformat:Half(87808,196,14,1) -> Half(43904,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.014948 [TRT] Tactic: 0 Time: 0.00789 [TRT] Fastest Tactic: 0 Time: 0.00789 [TRT] *************** Autotuning Reformat:Half(43904,196:2,14,1) -> Float(87808,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0163025 [TRT] Tactic: 0 Time: 0.007916 [TRT] Fastest Tactic: 0 Time: 0.007916 [TRT] *************** Autotuning Reformat:Half(43904,196:2,14,1) -> Float(87808,1,6272,448) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0103125 [TRT] Tactic: 0 Time: 0.01487 [TRT] Fastest Tactic: 1002 Time: 0.0103125 [TRT] *************** Autotuning Reformat:Half(43904,196:2,14,1) -> Half(87808,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.024375 [TRT] Tactic: 0 Time: 0.0126825 [TRT] Fastest Tactic: 0 Time: 0.0126825 [TRT] *************** Autotuning format combination: Float(87808,196,14,1) -> Float(163072,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/5x5 + inception_4e/relu_5x5 (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4e/5x5 + inception_4e/relu_5x5 (FusedConvActConvolution) [TRT] Tactic: 393215 Time: 0.355286 [TRT] Tactic: 917503 Time: 0.298151 [TRT] Tactic: 1114111 Time: 0.32349 [TRT] Tactic: 1245183 Time: 0.304245 [TRT] Tactic: 1572863 Time: 0.291641 [TRT] Tactic: 2490367 Time: 0.39664 [TRT] Tactic: 2555903 Time: 0.358594 [TRT] Tactic: 2949119 Time: 0.360104 [TRT] Tactic: 3211263 Time: 0.34448 [TRT] Tactic: 3801087 Time: 0.371197 [TRT] Tactic: 3866623 Time: 0.336511 [TRT] Tactic: 4128767 Time: 0.324687 [TRT] Tactic: 4456447 Time: 0.294167 [TRT] Tactic: 4718591 Time: 0.329765 [TRT] Tactic: 4784127 Time: 0.415521 [TRT] Tactic: 4849663 Time: 0.292344 [TRT] Tactic: 5111807 Time: 0.310677 [TRT] Tactic: 5308415 Time: 0.371718 [TRT] Tactic: 5505023 Time: 0.349818 [TRT] Tactic: 6094847 Time: 0.364115 [TRT] Tactic: 6356991 Time: 0.366381 [TRT] Tactic: 6553599 Time: 0.35586 [TRT] Tactic: 6619135 Time: 0.373073 [TRT] Tactic: 6684671 Time: 0.390078 [TRT] Tactic: 7471103 Time: 0.338281 [TRT] Tactic: 7667711 Time: 0.32362 [TRT] Tactic: 7929855 Time: 0.446562 [TRT] Tactic: 8060927 Time: 0.438568 [TRT] Tactic: 8126463 Time: 0.388281 [TRT] Tactic: 8388607 Time: 0.422083 [TRT] Tactic: 8519679 Time: 0.542995 [TRT] Tactic: 8781823 Time: 0.362291 [TRT] Tactic: 8912895 Time: 0.400105 [TRT] Tactic: 9240575 Time: 0.390938 [TRT] Tactic: 9306111 Time: 0.496276 [TRT] Tactic: 9371647 Time: 0.324244 [TRT] Tactic: 9437183 Time: 0.358411 [TRT] Tactic: 9633791 Time: 0.29336 [TRT] Tactic: 9699327 Time: 0.41362 [TRT] Tactic: 9764863 Time: 0.369922 [TRT] Tactic: 10158079 Time: 0.391406 [TRT] Tactic: 10420223 Time: 0.430313 [TRT] Tactic: 10616831 Time: 0.36763 [TRT] Tactic: 10878975 Time: 0.380573 [TRT] Fastest Tactic: 1572863 Time: 0.291641 [TRT] --------------- Timing Runner: inception_4e/5x5 + inception_4e/relu_5x5 (CudnnConvolution) [TRT] Tactic: 0 Time: 0.633307 [TRT] Tactic: 1 Time: 0.634426 [TRT] Tactic: 2 Time: 0.589974 [TRT] Tactic: 4 skipped. Scratch requested: 36339712, available: 33554432 [TRT] Tactic: 5 Time: 6.10247 [TRT] Fastest Tactic: 2 Time: 0.589974 [TRT] --------------- Timing Runner: inception_4e/5x5 + inception_4e/relu_5x5 (CaskConvolution) [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.488177 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v0 Tactic: 1754984623894446479 [TRT] Tactic: 1754984623894446479 Time: 0.503359 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v0 Tactic: 3611739942397549984 [TRT] Tactic: 3611739942397549984 Time: 0.293723 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v1 Tactic: 4337000649858996379 [TRT] Tactic: 4337000649858996379 Time: 0.312266 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.295443 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.284505 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.277265 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.308619 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.369713 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v1 Tactic: -9137461792520977713 [TRT] Tactic: -9137461792520977713 Time: 0.299141 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.298099 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v0 Tactic: -8133971918129952780 [TRT] Tactic: -8133971918129952780 Time: 0.322031 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v1 Tactic: -6092040395344634144 [TRT] Tactic: -6092040395344634144 Time: 0.506145 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.502421 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.352735 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.319401 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.296824 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.273672 [TRT] Fastest Tactic: -410470605513481746 Time: 0.273672 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -410470605513481746 [TRT] *************** Autotuning format combination: Float(87808,1,6272,448) -> Float(163072,1,11648,832) *************** [TRT] --------------- Timing Runner: inception_4e/5x5 + inception_4e/relu_5x5 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4e/5x5 + inception_4e/relu_5x5 (CaskConvolution) [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.352787 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.274817 [TRT] Fastest Tactic: -7394439838318485025 Time: 0.274817 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(87808,196,14,1) -> Half(163072,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/5x5 + inception_4e/relu_5x5 (CudnnConvolution) [TRT] Tactic: 0 Time: 0.655443 [TRT] Tactic: 1 Time: 0.639557 [TRT] Tactic: 2 Time: 0.585521 [TRT] Tactic: 4 skipped. Scratch requested: 36339712, available: 33554432 [TRT] Tactic: 5 Time: 6.25102 [TRT] Fastest Tactic: 2 Time: 0.585521 [TRT] --------------- Timing Runner: inception_4e/5x5 + inception_4e/relu_5x5 (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 2 [TRT] *************** Autotuning format combination: Half(43904,196:2,14,1) -> Half(81536,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/5x5 + inception_4e/relu_5x5 (FusedConvActConvolution) [TRT] Tactic: 393215 Time: 0.182474 [TRT] Tactic: 917503 Time: 0.168541 [TRT] Tactic: 1114111 Time: 0.165365 [TRT] Tactic: 1245183 Time: 0.167917 [TRT] Tactic: 1572863 Time: 0.169219 [TRT] Tactic: 2490367 Time: 0.242761 [TRT] Tactic: 2555903 Time: 0.231458 [TRT] Tactic: 2949119 Time: 0.174401 [TRT] Tactic: 3211263 Time: 0.177682 [TRT] Tactic: 3801087 Time: 0.173984 [TRT] Tactic: 3866623 Time: 0.175182 [TRT] Tactic: 4128767 Time: 0.15974 [TRT] Tactic: 4456447 Time: 0.167136 [TRT] Tactic: 4718591 Time: 0.14526 [TRT] Tactic: 4784127 Time: 0.21513 [TRT] Tactic: 4849663 Time: 0.15 [TRT] Tactic: 5111807 Time: 0.153464 [TRT] Tactic: 5308415 Time: 0.18263 [TRT] Tactic: 5505023 Time: 0.158125 [TRT] Tactic: 6094847 Time: 0.176068 [TRT] Tactic: 6356991 Time: 0.226198 [TRT] Tactic: 6553599 Time: 0.181745 [TRT] Tactic: 6619135 Time: 0.186485 [TRT] Tactic: 6684671 Time: 0.169843 [TRT] Tactic: 7471103 Time: 0.178567 [TRT] Tactic: 7667711 Time: 0.144557 [TRT] Tactic: 7929855 Time: 0.160313 [TRT] Tactic: 8060927 Time: 0.205105 [TRT] Tactic: 8126463 Time: 0.159218 [TRT] Tactic: 8388607 Time: 0.202266 [TRT] Tactic: 8519679 Time: 0.204818 [TRT] Tactic: 8781823 Time: 0.18 [TRT] Tactic: 8912895 Time: 0.176224 [TRT] Tactic: 9240575 Time: 0.143568 [TRT] Tactic: 9306111 Time: 0.24823 [TRT] Tactic: 9371647 Time: 0.160782 [TRT] Tactic: 9437183 Time: 0.174583 [TRT] Tactic: 9633791 Time: 0.150313 [TRT] Tactic: 9699327 Time: 0.1869 [TRT] Tactic: 9764863 Time: 0.202969 [TRT] Tactic: 10158079 Time: 0.188881 [TRT] Tactic: 10420223 Time: 0.194739 [TRT] Tactic: 10616831 Time: 0.173489 [TRT] Tactic: 10878975 Time: 0.207526 [TRT] Fastest Tactic: 9240575 Time: 0.143568 [TRT] --------------- Timing Runner: inception_4e/5x5 + inception_4e/relu_5x5 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4e/5x5 + inception_4e/relu_5x5 (CaskConvolution) [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.24961 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_large_nn_v1 Tactic: 3650389455493082349 [TRT] Tactic: 3650389455493082349 Time: 0.256823 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.180234 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.17151 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_large_nn_v1 Tactic: -6490690591794140522 [TRT] Tactic: -6490690591794140522 Time: 0.172734 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_large_nn_v1 Tactic: -4686027666808657977 [TRT] Tactic: -4686027666808657977 Time: 0.155677 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.142552 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.154687 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.150781 [TRT] Fastest Tactic: -4212163711445252890 Time: 0.142552 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -4212163711445252890 [TRT] *************** Autotuning Reformat:Float(163072,196,14,1) -> Float(163072,1,11648,832) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.03526 [TRT] Tactic: 0 Time: 0.0410155 [TRT] Fastest Tactic: 1002 Time: 0.03526 [TRT] *************** Autotuning Reformat:Float(163072,196,14,1) -> Half(163072,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.819922 [TRT] Tactic: 0 Time: 0.038125 [TRT] Fastest Tactic: 0 Time: 0.038125 [TRT] *************** Autotuning Reformat:Float(163072,196,14,1) -> Half(81536,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0451305 [TRT] Tactic: 0 Time: 0.0240365 [TRT] Fastest Tactic: 0 Time: 0.0240365 [TRT] *************** Autotuning Reformat:Float(163072,1,11648,832) -> Float(163072,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.046302 [TRT] Tactic: 0 Time: 0.0381245 [TRT] Fastest Tactic: 0 Time: 0.0381245 [TRT] *************** Autotuning Reformat:Float(163072,1,11648,832) -> Half(163072,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.033359 [TRT] Tactic: 0 Time: 0.038021 [TRT] Fastest Tactic: 1002 Time: 0.033359 [TRT] *************** Autotuning Reformat:Float(163072,1,11648,832) -> Half(81536,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0404165 [TRT] Tactic: 0 Time: 0.045078 [TRT] Fastest Tactic: 1002 Time: 0.0404165 [TRT] *************** Autotuning Reformat:Half(163072,196,14,1) -> Float(163072,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.839583 [TRT] Tactic: 0 Time: 0.038151 [TRT] Fastest Tactic: 0 Time: 0.038151 [TRT] *************** Autotuning Reformat:Half(163072,196,14,1) -> Float(163072,1,11648,832) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.031406 [TRT] Tactic: 0 Time: 0.0428385 [TRT] Fastest Tactic: 1002 Time: 0.031406 [TRT] *************** Autotuning Reformat:Half(163072,196,14,1) -> Half(81536,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0348695 [TRT] Tactic: 0 Time: 0.0241925 [TRT] Fastest Tactic: 0 Time: 0.0241925 [TRT] *************** Autotuning Reformat:Half(81536,196:2,14,1) -> Float(163072,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.045078 [TRT] Tactic: 0 Time: 0.045026 [TRT] Fastest Tactic: 0 Time: 0.045026 [TRT] *************** Autotuning Reformat:Half(81536,196:2,14,1) -> Float(163072,1,11648,832) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.031406 [TRT] Tactic: 0 Time: 0.047526 [TRT] Fastest Tactic: 1002 Time: 0.031406 [TRT] *************** Autotuning Reformat:Half(81536,196:2,14,1) -> Half(163072,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.057839 [TRT] Tactic: 0 Time: 0.040391 [TRT] Fastest Tactic: 0 Time: 0.040391 [TRT] *************** Autotuning Reformat:Float(103488,196,14,1) -> Half(103488,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(103488,196,14,1) -> Half(51744,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Float(103488,1,7392,528) -> Float(103488,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(103488,1,7392,528) -> Half(103488,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(103488,1,7392,528) -> Half(51744,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(103488,196,14,1) -> Float(103488,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(103488,196,14,1) -> Half(51744,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(51744,196:2,14,1) -> Float(103488,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(51744,196:2,14,1) -> Half(103488,196,14,1) *************** [TRT] *************** Autotuning format combination: Float(103488,196,14,1) -> Float(103488,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/pool (TiledPooling) [TRT] Tactic: 2752769 Time: 0.220131 [TRT] Tactic: 2818305 Time: 0.208958 [TRT] Tactic: 2883841 Time: 0.156094 [TRT] Tactic: 2949377 Time: 0.847994 [TRT] Tactic: 3014913 Time: 0.765728 [TRT] Tactic: 3080449 Time: 0.42039 [TRT] Tactic: 3145985 Time: 0.340156 [TRT] Tactic: 3211521 Time: 0.141692 [TRT] Tactic: 3277057 Time: 0.136849 [TRT] Tactic: 3342593 Time: 0.105338 [TRT] Tactic: 3408129 Time: 0.492292 [TRT] Tactic: 3473665 Time: 0.444297 [TRT] Tactic: 3539201 Time: 0.258204 [TRT] Tactic: 3604737 Time: 0.207839 [TRT] Tactic: 3670273 Time: 0.114505 [TRT] Tactic: 3735809 Time: 0.112239 [TRT] Tactic: 3801345 Time: 0.092604 [TRT] Tactic: 3866881 Time: 0.386745 [TRT] Tactic: 3932417 Time: 0.341719 [TRT] Tactic: 3997953 Time: 0.209557 [TRT] Tactic: 4063489 Time: 0.169349 [TRT] Tactic: 4129025 Time: 0.112422 [TRT] Tactic: 4194561 Time: 0.110078 [TRT] Tactic: 4260097 Time: 0.0874215 [TRT] Tactic: 4325633 Time: 0.330703 [TRT] Tactic: 4391169 Time: 0.292552 [TRT] Tactic: 4456705 Time: 0.184974 [TRT] Tactic: 4522241 Time: 0.150365 [TRT] Tactic: 4587777 Time: 0.112969 [TRT] Tactic: 4653313 Time: 0.110027 [TRT] Tactic: 4718849 Time: 0.083594 [TRT] Tactic: 4784385 Time: 0.306823 [TRT] Tactic: 4849921 Time: 0.276015 [TRT] Tactic: 4915457 Time: 0.174401 [TRT] Tactic: 4980993 Time: 0.140078 [TRT] Tactic: 5046529 Time: 0.114401 [TRT] Tactic: 5112065 Time: 0.113438 [TRT] Tactic: 5177601 Time: 0.0834375 [TRT] Tactic: 5243137 Time: 0.28539 [TRT] Tactic: 5308673 Time: 0.255651 [TRT] Tactic: 5374209 Time: 0.167552 [TRT] Tactic: 5439745 Time: 0.132474 [TRT] Tactic: 6553857 Time: 0.165521 [TRT] Tactic: 6750465 Time: 0.112005 [TRT] Fastest Tactic: 5177601 Time: 0.0834375 [TRT] --------------- Timing Runner: inception_4e/pool (CudnnPooling) [TRT] Tactic: -1 Time: 0.331588 [TRT] Fastest Tactic: -1 Time: 0.331588 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: TiledPooling Tactic: 5177601 [TRT] *************** Autotuning format combination: Half(103488,196,14,1) -> Half(103488,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/pool (TiledPooling) [TRT] TiledPooling has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4e/pool (CudnnPooling) [TRT] Tactic: -1 Time: 0.345833 [TRT] Fastest Tactic: -1 Time: 0.345833 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnPooling Tactic: -1 [TRT] *************** Autotuning format combination: Half(51744,196:2,14,1) -> Half(51744,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/pool (TiledPooling) [TRT] Tactic: 2752769 Time: 0.124062 [TRT] Tactic: 2818305 Time: 0.124141 [TRT] Tactic: 2883841 Time: 0.09362 [TRT] Tactic: 2949377 Time: 0.462214 [TRT] Tactic: 3014913 Time: 0.432344 [TRT] Tactic: 3080449 Time: 0.243046 [TRT] Tactic: 3145985 Time: 0.191198 [TRT] Tactic: 3211521 Time: 0.089115 [TRT] Tactic: 3277057 Time: 0.088854 [TRT] Tactic: 3342593 Time: 0.0684635 [TRT] Tactic: 3408129 Time: 0.309245 [TRT] Tactic: 3473665 Time: 0.281198 [TRT] Tactic: 3539201 Time: 0.163047 [TRT] Tactic: 3604737 Time: 0.135781 [TRT] Tactic: 3670273 Time: 0.077161 [TRT] Tactic: 3735809 Time: 0.074844 [TRT] Tactic: 3801345 Time: 0.0600255 [TRT] Tactic: 3866881 Time: 0.26375 [TRT] Tactic: 3932417 Time: 0.238751 [TRT] Tactic: 3997953 Time: 0.14138 [TRT] Tactic: 4063489 Time: 0.11888 [TRT] Tactic: 4129025 Time: 0.06862 [TRT] Tactic: 4194561 Time: 0.0695835 [TRT] Tactic: 4260097 Time: 0.0578915 [TRT] Tactic: 4325633 Time: 0.232917 [TRT] Tactic: 4391169 Time: 0.218282 [TRT] Tactic: 4456705 Time: 0.130807 [TRT] Tactic: 4522241 Time: 0.108464 [TRT] Tactic: 4587777 Time: 0.068281 [TRT] Tactic: 4653313 Time: 0.067708 [TRT] Tactic: 4718849 Time: 0.056354 [TRT] Tactic: 4784385 Time: 0.223646 [TRT] Tactic: 4849921 Time: 0.206693 [TRT] Tactic: 4915457 Time: 0.125234 [TRT] Tactic: 4980993 Time: 0.105912 [TRT] Tactic: 5046529 Time: 0.0654685 [TRT] Tactic: 5112065 Time: 0.066172 [TRT] Tactic: 5177601 Time: 0.05487 [TRT] Tactic: 5243137 Time: 0.215912 [TRT] Tactic: 5308673 Time: 0.196901 [TRT] Tactic: 5374209 Time: 0.119714 [TRT] Tactic: 5439745 Time: 0.102734 [TRT] Tactic: 6553857 Time: 0.105651 [TRT] Tactic: 6750465 Time: 0.0760935 [TRT] Fastest Tactic: 5177601 Time: 0.05487 [TRT] --------------- Timing Runner: inception_4e/pool (CudaPooling) [TRT] Tactic: -3 Time: 0.191197 [TRT] Fastest Tactic: -3 Time: 0.191197 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: TiledPooling Tactic: 5177601 [TRT] *************** Autotuning Reformat:Float(103488,196,14,1) -> Float(103488,1,7392,528) *************** [TRT] *************** Autotuning Reformat:Float(103488,196,14,1) -> Half(103488,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(103488,196,14,1) -> Half(51744,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Float(103488,1,7392,528) -> Float(103488,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(103488,1,7392,528) -> Half(103488,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(103488,1,7392,528) -> Half(51744,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(103488,196,14,1) -> Float(103488,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(103488,196,14,1) -> Float(103488,1,7392,528) *************** [TRT] *************** Autotuning Reformat:Half(103488,196,14,1) -> Half(51744,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(51744,196:2,14,1) -> Float(103488,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(51744,196:2,14,1) -> Float(103488,1,7392,528) *************** [TRT] *************** Autotuning Reformat:Half(51744,196:2,14,1) -> Half(103488,196,14,1) *************** [TRT] *************** Autotuning format combination: Float(103488,196,14,1) -> Float(163072,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/pool_proj + inception_4e/relu_pool_proj (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4e/pool_proj + inception_4e/relu_pool_proj (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.500234 [TRT] Tactic: 655359 Time: 0.270625 [TRT] Tactic: 786431 Time: 0.409948 [TRT] Tactic: 851967 Time: 0.400364 [TRT] Tactic: 1179647 Time: 0.433516 [TRT] Tactic: 1310719 Time: 0.979401 [TRT] Tactic: 1376255 Time: 0.296797 [TRT] Tactic: 1441791 Time: 0.522761 [TRT] Tactic: 1507327 Time: 0.494505 [TRT] Tactic: 1638399 Time: 0.603307 [TRT] Tactic: 1835007 Time: 0.432552 [TRT] Tactic: 1900543 Time: 0.476198 [TRT] Tactic: 2162687 Time: 0.301328 [TRT] Tactic: 2293759 Time: 0.315287 [TRT] Tactic: 2359295 Time: 0.423281 [TRT] Tactic: 2686975 Time: 0.432526 [TRT] Tactic: 3080191 Time: 0.334244 [TRT] Tactic: 3342335 Time: 0.388828 [TRT] Tactic: 3407871 Time: 0.326719 [TRT] Tactic: 3538943 Time: 0.348386 [TRT] Tactic: 3670015 Time: 0.296822 [TRT] Tactic: 3932159 Time: 0.423855 [TRT] Tactic: 3997695 Time: 0.42138 [TRT] Tactic: 4063231 Time: 0.345859 [TRT] Tactic: 4194303 Time: 0.428308 [TRT] Tactic: 4325375 Time: 0.496459 [TRT] Tactic: 4521983 Time: 0.539791 [TRT] Tactic: 4587519 Time: 0.454636 [TRT] Tactic: 4653055 Time: 0.436303 [TRT] Tactic: 4915199 Time: 0.399896 [TRT] Tactic: 4980735 Time: 0.523438 [TRT] Tactic: 5177343 Time: 0.50526 [TRT] Tactic: 5242879 Time: 0.292396 [TRT] Tactic: 5373951 Time: 0.483333 [TRT] Tactic: 5439487 Time: 0.425234 [TRT] Tactic: 5570559 Time: 0.278723 [TRT] Tactic: 5636095 Time: 0.351067 [TRT] Tactic: 5701631 Time: 0.445625 [TRT] Tactic: 5767167 Time: 0.588151 [TRT] Tactic: 5832703 Time: 0.323073 [TRT] Tactic: 5898239 Time: 0.325078 [TRT] Tactic: 6029311 Time: 0.294479 [TRT] Tactic: 6225919 Time: 0.310729 [TRT] Tactic: 6291455 Time: 0.430937 [TRT] Tactic: 6422527 Time: 0.337005 [TRT] Tactic: 6750207 Time: 0.381719 [TRT] Tactic: 6815743 Time: 0.302604 [TRT] Tactic: 6946815 Time: 0.513829 [TRT] Tactic: 7077887 Time: 0.346979 [TRT] Tactic: 7143423 Time: 0.620755 [TRT] Tactic: 7208959 Time: 0.402656 [TRT] Tactic: 7340031 Time: 0.346042 [TRT] Tactic: 7405567 Time: 0.351458 [TRT] Tactic: 7536639 Time: 0.339427 [TRT] Tactic: 7602175 Time: 0.483854 [TRT] Tactic: 7733247 Time: 0.325989 [TRT] Tactic: 7798783 Time: 0.410599 [TRT] Tactic: 8191999 Time: 0.61763 [TRT] Tactic: 8257535 Time: 0.409739 [TRT] Tactic: 8323071 Time: 0.399426 [TRT] Tactic: 8650751 Time: 0.5775 [TRT] Tactic: 8716287 Time: 0.334818 [TRT] Tactic: 9568255 Time: 0.399193 [TRT] Tactic: 9895935 Time: 0.428126 [TRT] Tactic: 10223615 Time: 0.433151 [TRT] Tactic: 10354687 Time: 0.491224 [TRT] Tactic: 10551295 Time: 0.354322 [TRT] Tactic: 10747903 Time: 0.314687 [TRT] Tactic: 10944511 Time: 0.52375 [TRT] Fastest Tactic: 655359 Time: 0.270625 [TRT] --------------- Timing Runner: inception_4e/pool_proj + inception_4e/relu_pool_proj (CudnnConvolution) [TRT] Tactic: 0 Time: 0.425365 [TRT] Tactic: 1 Time: 0.357605 [TRT] Tactic: 2 Time: 0.55237 [TRT] Tactic: 4 skipped. Scratch requested: 157200384, available: 33554432 [TRT] Tactic: 5 Time: 4.89737 [TRT] Fastest Tactic: 1 Time: 0.357605 [TRT] --------------- Timing Runner: inception_4e/pool_proj + inception_4e/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4e/pool_proj + inception_4e/relu_pool_proj (CaskConvolution) [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.276224 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v0 Tactic: 1698681053543049347 [TRT] Tactic: 1698681053543049347 Time: 0.269792 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.189792 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.200703 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.191381 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v1 Tactic: 5326823351883942011 [TRT] Tactic: 5326823351883942011 Time: 0.185 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v0 Tactic: 5500448035057547314 [TRT] Tactic: 5500448035057547314 Time: 0.210781 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.205859 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.278854 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.192995 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v1 Tactic: -6576203419454146580 [TRT] Tactic: -6576203419454146580 Time: 0.24112 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.291823 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.252578 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.219687 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.211927 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.187084 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v0 Tactic: -377491875521947884 [TRT] Tactic: -377491875521947884 Time: 0.187734 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v1 Tactic: -37215280111360163 [TRT] Tactic: -37215280111360163 Time: 0.194479 [TRT] Fastest Tactic: 5326823351883942011 Time: 0.185 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 5326823351883942011 [TRT] *************** Autotuning format combination: Float(103488,1,7392,528) -> Float(163072,1,11648,832) *************** [TRT] --------------- Timing Runner: inception_4e/pool_proj + inception_4e/relu_pool_proj (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4e/pool_proj + inception_4e/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4e/pool_proj + inception_4e/relu_pool_proj (CaskConvolution) [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 3886731678879822788 [TRT] Tactic: 3886731678879822788 Time: 0.20138 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 6629944304117643200 [TRT] Tactic: 6629944304117643200 Time: 0.320886 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.323932 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.198724 [TRT] Fastest Tactic: -7394439838318485025 Time: 0.198724 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(103488,196,14,1) -> Half(163072,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/pool_proj + inception_4e/relu_pool_proj (CudnnConvolution) [TRT] Tactic: 0 Time: 0.46474 [TRT] Tactic: 1 Time: 0.392265 [TRT] Tactic: 2 Time: 0.541901 [TRT] Tactic: 4 skipped. Scratch requested: 157200384, available: 33554432 [TRT] Tactic: 5 Time: 4.7955 [TRT] Fastest Tactic: 1 Time: 0.392265 [TRT] --------------- Timing Runner: inception_4e/pool_proj + inception_4e/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4e/pool_proj + inception_4e/relu_pool_proj (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 1 [TRT] *************** Autotuning format combination: Half(51744,196:2,14,1) -> Half(163072,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/pool_proj + inception_4e/relu_pool_proj (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] *************** Autotuning format combination: Half(51744,196:2,14,1) -> Half(81536,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/pool_proj + inception_4e/relu_pool_proj (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.230599 [TRT] Tactic: 655359 Time: 0.235052 [TRT] Tactic: 786431 Time: 0.259323 [TRT] Tactic: 851967 Time: 0.206849 [TRT] Tactic: 1310719 Time: 0.521719 [TRT] Tactic: 1376255 Time: 0.156458 [TRT] Tactic: 1507327 Time: 0.248385 [TRT] Tactic: 1638399 Time: 0.297265 [TRT] Tactic: 1835007 Time: 0.25487 [TRT] Tactic: 1900543 Time: 0.243515 [TRT] Tactic: 2162687 Time: 0.238698 [TRT] Tactic: 2293759 Time: 0.214167 [TRT] Tactic: 2359295 Time: 0.208646 [TRT] Tactic: 2686975 Time: 0.385443 [TRT] Tactic: 3080191 Time: 0.181172 [TRT] Tactic: 3342335 Time: 0.205651 [TRT] Tactic: 3407871 Time: 0.175703 [TRT] Tactic: 3538943 Time: 0.170963 [TRT] Tactic: 3670015 Time: 0.236693 [TRT] Tactic: 3932159 Time: 0.184532 [TRT] Tactic: 4063231 Time: 0.175756 [TRT] Tactic: 4194303 Time: 0.212812 [TRT] Tactic: 4325375 Time: 0.256433 [TRT] Tactic: 4521983 Time: 0.261693 [TRT] Tactic: 4980735 Time: 0.255885 [TRT] Tactic: 5242879 Time: 0.151015 [TRT] Tactic: 5439487 Time: 0.217187 [TRT] Tactic: 5570559 Time: 0.19625 [TRT] Tactic: 5636095 Time: 0.175912 [TRT] Tactic: 5701631 Time: 0.200104 [TRT] Tactic: 5767167 Time: 0.290443 [TRT] Tactic: 5832703 Time: 0.163854 [TRT] Tactic: 6029311 Time: 0.210077 [TRT] Tactic: 6225919 Time: 0.147527 [TRT] Tactic: 6422527 Time: 0.169323 [TRT] Tactic: 6815743 Time: 0.160261 [TRT] Tactic: 6946815 Time: 0.2469 [TRT] Tactic: 7077887 Time: 0.161589 [TRT] Tactic: 7143423 Time: 0.311067 [TRT] Tactic: 7208959 Time: 0.187187 [TRT] Tactic: 7405567 Time: 0.174245 [TRT] Tactic: 7536639 Time: 0.20987 [TRT] Tactic: 7602175 Time: 0.227344 [TRT] Tactic: 7733247 Time: 0.181718 [TRT] Tactic: 7798783 Time: 0.257344 [TRT] Tactic: 8191999 Time: 0.308984 [TRT] Tactic: 8323071 Time: 0.205156 [TRT] Tactic: 8650751 Time: 0.265572 [TRT] Tactic: 8716287 Time: 0.163047 [TRT] Tactic: 9895935 Time: 0.212813 [TRT] Tactic: 10223615 Time: 0.377604 [TRT] Tactic: 10551295 Time: 0.185964 [TRT] Tactic: 10747903 Time: 0.164375 [TRT] Tactic: 10944511 Time: 0.256536 [TRT] Fastest Tactic: 6225919 Time: 0.147527 [TRT] --------------- Timing Runner: inception_4e/pool_proj + inception_4e/relu_pool_proj (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4e/pool_proj + inception_4e/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_4e/pool_proj + inception_4e/relu_pool_proj (CaskConvolution) [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_interior_nn_v1 Tactic: 3066127711859985668 [TRT] Tactic: 3066127711859985668 Time: 0.126875 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.146224 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.132499 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.110052 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] Tactic: 8163473458334948789 Time: 0.102005 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.0996875 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.10263 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.106876 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_interior_nn_v1 Tactic: -1716393687483585322 [TRT] Tactic: -1716393687483585322 Time: 0.099609 [TRT] Fastest Tactic: -1716393687483585322 Time: 0.099609 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -1716393687483585322 [TRT] *************** Autotuning Reformat:Float(163072,196,14,1) -> Float(163072,1,11648,832) *************** [TRT] *************** Autotuning Reformat:Float(163072,196,14,1) -> Half(163072,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(163072,196,14,1) -> Half(81536,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Float(163072,1,11648,832) -> Float(163072,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(163072,1,11648,832) -> Half(163072,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(163072,1,11648,832) -> Half(81536,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(163072,196,14,1) -> Float(163072,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(163072,196,14,1) -> Float(163072,1,11648,832) *************** [TRT] *************** Autotuning Reformat:Half(163072,196,14,1) -> Half(81536,196:2,14,1) *************** [TRT] *************** Autotuning Reformat:Half(81536,196:2,14,1) -> Float(163072,196,14,1) *************** [TRT] *************** Autotuning Reformat:Half(81536,196:2,14,1) -> Float(163072,1,11648,832) *************** [TRT] *************** Autotuning Reformat:Half(81536,196:2,14,1) -> Half(163072,196,14,1) *************** [TRT] *************** Autotuning Reformat:Float(87808,196,14,1) -> Float(163072,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 3.75503 [TRT] Tactic: 0 Time: 0.029219 [TRT] Fastest Tactic: 0 Time: 0.029219 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(87808,196,14,1) -> Float(163072,1,11648,832) *************** [TRT] --------------- Timing Runner: inception_4e/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.073073 [TRT] Tactic: 0 Time: 0.082526 [TRT] Fastest Tactic: 1002 Time: 0.073073 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(87808,196,14,1) -> Half(163072,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 1.63224 [TRT] Tactic: 0 Time: 0.073802 [TRT] Fastest Tactic: 0 Time: 0.073802 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(87808,196,14,1) -> Half(81536,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0861455 [TRT] Tactic: 0 Time: 0.0878125 [TRT] Fastest Tactic: 1002 Time: 0.0861455 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(87808,1,6272,448) -> Float(163072,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0951565 [TRT] Tactic: 0 Time: 0.0779425 [TRT] Fastest Tactic: 0 Time: 0.0779425 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(87808,1,6272,448) -> Float(163072,1,11648,832) *************** [TRT] --------------- Timing Runner: inception_4e/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0670315 [TRT] Tactic: 0 Time: 0.0785425 [TRT] Fastest Tactic: 1002 Time: 0.0670315 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(87808,1,6272,448) -> Half(163072,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.061953 [TRT] Tactic: 0 Time: 0.0757815 [TRT] Fastest Tactic: 1002 Time: 0.061953 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(87808,1,6272,448) -> Half(81536,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0742705 [TRT] Tactic: 0 Time: 0.0877865 [TRT] Fastest Tactic: 1002 Time: 0.0742705 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(87808,196,14,1) -> Float(163072,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 1.67034 [TRT] Tactic: 0 Time: 0.075729 [TRT] Fastest Tactic: 0 Time: 0.075729 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(87808,196,14,1) -> Float(163072,1,11648,832) *************** [TRT] --------------- Timing Runner: inception_4e/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.056901 [TRT] Tactic: 0 Time: 0.080885 [TRT] Fastest Tactic: 1002 Time: 0.056901 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(87808,196,14,1) -> Half(163072,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 1.66458 [TRT] Tactic: 0 Time: 0.0102865 [TRT] Fastest Tactic: 0 Time: 0.0102865 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(87808,196,14,1) -> Half(81536,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.060833 [TRT] Tactic: 0 Time: 0.045104 [TRT] Fastest Tactic: 0 Time: 0.045104 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(43904,196:2,14,1) -> Float(163072,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0813025 [TRT] Tactic: 0 Time: 0.0868225 [TRT] Fastest Tactic: 1002 Time: 0.0813025 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(43904,196:2,14,1) -> Float(163072,1,11648,832) *************** [TRT] --------------- Timing Runner: inception_4e/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.057213 [TRT] Tactic: 0 Time: 0.093411 [TRT] Fastest Tactic: 1002 Time: 0.057213 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(43904,196:2,14,1) -> Half(163072,196,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.108906 [TRT] Tactic: 0 Time: 0.0778905 [TRT] Fastest Tactic: 0 Time: 0.0778905 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(43904,196:2,14,1) -> Half(81536,196:2,14,1) *************** [TRT] --------------- Timing Runner: inception_4e/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0824215 [TRT] Tactic: 0 Time: 0.010286 [TRT] Fastest Tactic: 0 Time: 0.010286 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(163072,196,14,1) -> Half(163072,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 4.38016 [TRT] Tactic: 0 Time: 0.173386 [TRT] Fastest Tactic: 0 Time: 0.173386 [TRT] *************** Autotuning Reformat:Float(163072,196,14,1) -> Half(81536,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.265573 [TRT] Tactic: 0 Time: 0.137006 [TRT] Fastest Tactic: 0 Time: 0.137006 [TRT] *************** Autotuning Reformat:Float(163072,1,11648,832) -> Float(163072,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.265261 [TRT] Tactic: 0 Time: 0.240104 [TRT] Fastest Tactic: 0 Time: 0.240104 [TRT] *************** Autotuning Reformat:Float(163072,1,11648,832) -> Half(163072,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.184141 [TRT] Tactic: 0 Time: 0.234453 [TRT] Fastest Tactic: 1002 Time: 0.184141 [TRT] *************** Autotuning Reformat:Float(163072,1,11648,832) -> Half(81536,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.230026 [TRT] Tactic: 0 Time: 0.273255 [TRT] Fastest Tactic: 1002 Time: 0.230026 [TRT] *************** Autotuning Reformat:Half(163072,196,14,1) -> Float(163072,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 4.51727 [TRT] Tactic: 0 Time: 0.146432 [TRT] Fastest Tactic: 0 Time: 0.146432 [TRT] *************** Autotuning Reformat:Half(163072,196,14,1) -> Half(81536,196:2,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.188542 [TRT] Tactic: 0 Time: 0.13612 [TRT] Fastest Tactic: 0 Time: 0.13612 [TRT] *************** Autotuning Reformat:Half(81536,196:2,14,1) -> Float(163072,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.240833 [TRT] Tactic: 0 Time: 0.12013 [TRT] Fastest Tactic: 0 Time: 0.12013 [TRT] *************** Autotuning Reformat:Half(81536,196:2,14,1) -> Half(163072,196,14,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.346927 [TRT] Tactic: 0 Time: 0.117136 [TRT] Fastest Tactic: 0 Time: 0.117136 [TRT] *************** Autotuning format combination: Float(163072,196,14,1) -> Float(40768,49,7,1) *************** [TRT] --------------- Timing Runner: pool4/3x3_s2 (TiledPooling) [TRT] Tactic: 257 Time: 0.194297 [TRT] Tactic: 65793 Time: 0.18823 [TRT] Tactic: 131329 Time: 0.824245 [TRT] Tactic: 196865 Time: 2.27224 [TRT] Tactic: 262401 Time: 1.70849 [TRT] Tactic: 327937 Time: 0.902474 [TRT] Tactic: 393473 Time: 0.825547 [TRT] Tactic: 459009 Time: 0.115286 [TRT] Tactic: 524545 Time: 0.112239 [TRT] Tactic: 590081 Time: 0.472396 [TRT] Tactic: 655617 Time: 1.33534 [TRT] Tactic: 721153 Time: 1.00758 [TRT] Tactic: 786689 Time: 0.539609 [TRT] Tactic: 852225 Time: 0.487214 [TRT] Tactic: 917761 Time: 0.0905205 [TRT] Tactic: 983297 Time: 0.0887245 [TRT] Tactic: 1048833 Time: 0.36112 [TRT] Tactic: 1114369 Time: 1.01297 [TRT] Tactic: 1179905 Time: 0.788099 [TRT] Tactic: 1245441 Time: 0.412161 [TRT] Tactic: 1310977 Time: 0.365157 [TRT] Tactic: 1376513 Time: 0.076719 [TRT] Tactic: 1442049 Time: 0.075912 [TRT] Tactic: 1507585 Time: 0.295364 [TRT] Tactic: 1573121 Time: 0.832109 [TRT] Tactic: 1638657 Time: 0.652135 [TRT] Tactic: 1704193 Time: 0.313724 [TRT] Tactic: 1769729 Time: 0.305156 [TRT] Tactic: 1835265 Time: 0.0698435 [TRT] Tactic: 1900801 Time: 0.0697395 [TRT] Tactic: 1966337 Time: 0.262447 [TRT] Tactic: 2031873 Time: 0.74375 [TRT] Tactic: 2097409 Time: 0.589948 [TRT] Tactic: 2162945 Time: 0.264713 [TRT] Tactic: 2228481 Time: 0.273125 [TRT] Tactic: 2294017 Time: 0.065729 [TRT] Tactic: 2359553 Time: 0.065208 [TRT] Tactic: 2425089 Time: 0.23914 [TRT] Tactic: 2490625 Time: 0.670312 [TRT] Tactic: 2556161 Time: 0.525677 [TRT] Tactic: 2621697 Time: 0.237552 [TRT] Tactic: 2687233 Time: 0.249896 [TRT] Tactic: 6947073 Time: 0.307187 [TRT] Fastest Tactic: 2359553 Time: 0.065208 [TRT] --------------- Timing Runner: pool4/3x3_s2 (CudnnPooling) [TRT] Tactic: -1 Time: 0.297162 [TRT] Fastest Tactic: -1 Time: 0.297162 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: TiledPooling Tactic: 2359553 [TRT] *************** Autotuning format combination: Half(163072,196,14,1) -> Half(40768,49,7,1) *************** [TRT] --------------- Timing Runner: pool4/3x3_s2 (TiledPooling) [TRT] TiledPooling has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: pool4/3x3_s2 (CudnnPooling) [TRT] Tactic: -1 Time: 0.307448 [TRT] Fastest Tactic: -1 Time: 0.307448 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnPooling Tactic: -1 [TRT] *************** Autotuning format combination: Half(81536,196:2,14,1) -> Half(20384,49:2,7,1) *************** [TRT] --------------- Timing Runner: pool4/3x3_s2 (TiledPooling) [TRT] Tactic: 257 Time: 0.104271 [TRT] Tactic: 65793 Time: 0.100495 [TRT] Tactic: 131329 Time: 0.433464 [TRT] Tactic: 196865 Time: 1.20672 [TRT] Tactic: 262401 Time: 0.910964 [TRT] Tactic: 327937 Time: 0.484844 [TRT] Tactic: 393473 Time: 0.442577 [TRT] Tactic: 459009 Time: 0.066224 [TRT] Tactic: 524545 Time: 0.0644535 [TRT] Tactic: 590081 Time: 0.272318 [TRT] Tactic: 655617 Time: 0.737448 [TRT] Tactic: 721153 Time: 0.577708 [TRT] Tactic: 786689 Time: 0.289219 [TRT] Tactic: 852225 Time: 0.274922 [TRT] Tactic: 917761 Time: 0.0527605 [TRT] Tactic: 983297 Time: 0.052474 [TRT] Tactic: 1048833 Time: 0.213828 [TRT] Tactic: 1114369 Time: 0.577109 [TRT] Tactic: 1179905 Time: 0.448385 [TRT] Tactic: 1245441 Time: 0.214114 [TRT] Tactic: 1310977 Time: 0.217579 [TRT] Tactic: 1376513 Time: 0.047292 [TRT] Tactic: 1442049 Time: 0.046354 [TRT] Tactic: 1507585 Time: 0.183697 [TRT] Tactic: 1573121 Time: 0.475442 [TRT] Tactic: 1638657 Time: 0.384062 [TRT] Tactic: 1704193 Time: 0.172422 [TRT] Tactic: 1769729 Time: 0.184688 [TRT] Tactic: 1835265 Time: 0.043698 [TRT] Tactic: 1900801 Time: 0.043932 [TRT] Tactic: 1966337 Time: 0.171979 [TRT] Tactic: 2031873 Time: 0.474401 [TRT] Tactic: 2097409 Time: 0.360469 [TRT] Tactic: 2162945 Time: 0.164661 [TRT] Tactic: 2228481 Time: 0.1725 [TRT] Tactic: 2294017 Time: 0.042786 [TRT] Tactic: 2359553 Time: 0.041979 [TRT] Tactic: 2425089 Time: 0.166536 [TRT] Tactic: 2490625 Time: 0.399037 [TRT] Tactic: 2556161 Time: 0.330807 [TRT] Tactic: 2621697 Time: 0.147604 [TRT] Tactic: 2687233 Time: 0.166328 [TRT] Tactic: 6947073 Time: 0.183906 [TRT] Fastest Tactic: 2359553 Time: 0.041979 [TRT] --------------- Timing Runner: pool4/3x3_s2 (CudaPooling) [TRT] Tactic: -3 Time: 0.0808335 [TRT] Fastest Tactic: -3 Time: 0.0808335 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: TiledPooling Tactic: 2359553 [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Float(40768,1,5824,832) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0604685 [TRT] Tactic: 0 Time: 0.0658595 [TRT] Fastest Tactic: 1002 Time: 0.0604685 [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Half(40768,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.25737 [TRT] Tactic: 0 Time: 0.04513 [TRT] Fastest Tactic: 0 Time: 0.04513 [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Half(20384,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.074505 [TRT] Tactic: 0 Time: 0.03724 [TRT] Fastest Tactic: 0 Time: 0.03724 [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Float(40768,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.2963 [TRT] Tactic: 0 Time: 0.0399745 [TRT] Fastest Tactic: 0 Time: 0.0399745 [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Float(40768,1,5824,832) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0541145 [TRT] Tactic: 0 Time: 0.0669535 [TRT] Fastest Tactic: 1002 Time: 0.0541145 [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Half(20384,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0803125 [TRT] Tactic: 0 Time: 0.0369265 [TRT] Fastest Tactic: 0 Time: 0.0369265 [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Float(40768,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.070521 [TRT] Tactic: 0 Time: 0.033021 [TRT] Fastest Tactic: 0 Time: 0.033021 [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Float(40768,1,5824,832) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0540625 [TRT] Tactic: 0 Time: 0.0760675 [TRT] Fastest Tactic: 1002 Time: 0.0540625 [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Half(40768,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.089375 [TRT] Tactic: 0 Time: 0.0310935 [TRT] Fastest Tactic: 0 Time: 0.0310935 [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Float(40768,1,5824,832) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.061198 [TRT] Tactic: 0 Time: 0.0660155 [TRT] Fastest Tactic: 1002 Time: 0.061198 [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Half(40768,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.25706 [TRT] Tactic: 0 Time: 0.044974 [TRT] Fastest Tactic: 0 Time: 0.044974 [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Half(20384,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0740365 [TRT] Tactic: 0 Time: 0.037083 [TRT] Fastest Tactic: 0 Time: 0.037083 [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Float(40768,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0781515 [TRT] Tactic: 0 Time: 0.060104 [TRT] Fastest Tactic: 0 Time: 0.060104 [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Half(40768,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.058125 [TRT] Tactic: 0 Time: 0.0604425 [TRT] Fastest Tactic: 1002 Time: 0.058125 [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Half(20384,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.06625 [TRT] Tactic: 0 Time: 0.072344 [TRT] Fastest Tactic: 1002 Time: 0.06625 [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Float(40768,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.29625 [TRT] Tactic: 0 Time: 0.03875 [TRT] Fastest Tactic: 0 Time: 0.03875 [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Float(40768,1,5824,832) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0530735 [TRT] Tactic: 0 Time: 0.066354 [TRT] Fastest Tactic: 1002 Time: 0.0530735 [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Half(20384,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0800525 [TRT] Tactic: 0 Time: 0.037266 [TRT] Fastest Tactic: 0 Time: 0.037266 [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Float(40768,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0698695 [TRT] Tactic: 0 Time: 0.0331515 [TRT] Fastest Tactic: 0 Time: 0.0331515 [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Float(40768,1,5824,832) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.053698 [TRT] Tactic: 0 Time: 0.076198 [TRT] Fastest Tactic: 1002 Time: 0.053698 [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Half(40768,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.089193 [TRT] Tactic: 0 Time: 0.031224 [TRT] Fastest Tactic: 0 Time: 0.031224 [TRT] *************** Autotuning format combination: Float(40768,49,7,1) -> Float(21952,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.486406 [TRT] Tactic: 655359 Time: 0.458176 [TRT] Tactic: 786431 Time: 0.918568 [TRT] Tactic: 851967 Time: 0.803334 [TRT] Tactic: 1179647 Time: 0.63586 [TRT] Tactic: 1310719 Time: 1.10143 [TRT] Tactic: 1376255 Time: 0.687604 [TRT] Tactic: 1441791 Time: 0.822734 [TRT] Tactic: 1507327 Time: 0.787084 [TRT] Tactic: 1638399 Time: 0.95224 [TRT] Tactic: 1835007 Time: 0.713516 [TRT] Tactic: 1900543 Time: 1.06599 [TRT] Tactic: 2097151 Time: 0.591953 [TRT] Tactic: 2162687 Time: 0.790234 [TRT] Tactic: 2293759 Time: 0.493724 [TRT] Tactic: 2359295 Time: 0.667344 [TRT] Tactic: 2686975 Time: 0.688073 [TRT] Tactic: 3080191 Time: 0.561484 [TRT] Tactic: 3342335 Time: 1.13479 [TRT] Tactic: 3407871 Time: 0.52513 [TRT] Tactic: 3538943 Time: 0.560234 [TRT] Tactic: 3670015 Time: 0.688985 [TRT] Tactic: 3932159 Time: 0.656407 [TRT] Tactic: 3997695 Time: 0.925183 [TRT] Tactic: 4063231 Time: 0.635287 [TRT] Tactic: 4194303 Time: 0.510677 [TRT] Tactic: 4259839 Time: 0.614219 [TRT] Tactic: 4325375 Time: 0.836172 [TRT] Tactic: 4521983 Time: 0.862213 [TRT] Tactic: 4587519 Time: 0.724167 [TRT] Tactic: 4653055 Time: 0.686328 [TRT] Tactic: 4915199 Time: 0.50625 [TRT] Tactic: 4980735 Time: 0.825442 [TRT] Tactic: 5177343 Time: 0.645625 [TRT] Tactic: 5242879 Time: 0.474245 [TRT] Tactic: 5373951 Time: 0.677786 [TRT] Tactic: 5439487 Time: 0.609792 [TRT] Tactic: 5570559 Time: 0.435053 [TRT] Tactic: 5636095 Time: 0.634479 [TRT] Tactic: 5701631 Time: 0.699167 [TRT] Tactic: 5767167 Time: 0.873255 [TRT] Tactic: 5832703 Time: 0.513619 [TRT] Tactic: 5898239 Time: 0.410729 [TRT] Tactic: 6029311 Time: 0.45961 [TRT] Tactic: 6225919 Time: 0.509792 [TRT] Tactic: 6291455 Time: 0.637604 [TRT] Tactic: 6422527 Time: 0.564766 [TRT] Tactic: 6750207 Time: 0.587448 [TRT] Tactic: 6815743 Time: 0.47776 [TRT] Tactic: 6946815 Time: 0.7906 [TRT] Tactic: 7012351 Time: 0.588881 [TRT] Tactic: 7077887 Time: 0.563698 [TRT] Tactic: 7143423 Time: 0.925417 [TRT] Tactic: 7208959 Time: 0.516484 [TRT] Tactic: 7340031 Time: 0.423698 [TRT] Tactic: 7405567 Time: 0.620338 [TRT] Tactic: 7536639 Time: 0.540521 [TRT] Tactic: 7602175 Time: 0.781718 [TRT] Tactic: 7733247 Time: 0.452812 [TRT] Tactic: 7798783 Time: 0.920859 [TRT] Tactic: 8191999 Time: 0.831953 [TRT] Tactic: 8257535 Time: 0.509349 [TRT] Tactic: 8323071 Time: 0.610521 [TRT] Tactic: 8650751 Time: 0.784088 [TRT] Tactic: 8716287 Time: 0.529244 [TRT] Tactic: 9109503 Time: 0.595286 [TRT] Tactic: 9568255 Time: 0.507214 [TRT] Tactic: 9895935 Time: 0.511536 [TRT] Tactic: 10223615 Time: 0.68586 [TRT] Tactic: 10354687 Time: 0.65263 [TRT] Tactic: 10551295 Time: 0.540704 [TRT] Tactic: 10747903 Time: 0.441588 [TRT] Tactic: 10944511 Time: 0.824349 [TRT] Fastest Tactic: 5898239 Time: 0.410729 [TRT] --------------- Timing Runner: inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce (CudnnConvolution) [TRT] Tactic: 0 Time: 0.66151 [TRT] Tactic: 1 Time: 0.567448 [TRT] Tactic: 2 Time: 0.770756 [TRT] Tactic: 4 skipped. Scratch requested: 862191616, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 51910656, available: 33554432 [TRT] Fastest Tactic: 1 Time: 0.567448 [TRT] --------------- Timing Runner: inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce (CaskConvolution) [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.682838 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v0 Tactic: 1698681053543049347 [TRT] Tactic: 1698681053543049347 Time: 0.6375 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.56961 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.499714 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.568594 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v1 Tactic: 5326823351883942011 [TRT] Tactic: 5326823351883942011 Time: 0.548151 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v0 Tactic: 5500448035057547314 [TRT] Tactic: 5500448035057547314 Time: 0.550235 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.509192 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.664218 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.580469 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v1 Tactic: -6576203419454146580 [TRT] Tactic: -6576203419454146580 Time: 0.597916 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.695495 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.628724 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.576511 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.566172 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.561302 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v0 Tactic: -377491875521947884 [TRT] Tactic: -377491875521947884 Time: 0.561641 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v1 Tactic: -37215280111360163 [TRT] Tactic: -37215280111360163 Time: 0.492135 [TRT] Fastest Tactic: -37215280111360163 Time: 0.492135 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 5898239 [TRT] *************** Autotuning format combination: Float(40768,1,5824,832) -> Float(21952,1,3136,448) *************** [TRT] --------------- Timing Runner: inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce (CaskConvolution) [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 3886731678879822788 [TRT] Tactic: 3886731678879822788 Time: 0.49599 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 6629944304117643200 [TRT] Tactic: 6629944304117643200 Time: 0.668568 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.66961 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.505911 [TRT] Fastest Tactic: 3886731678879822788 Time: 0.49599 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 3886731678879822788 [TRT] *************** Autotuning format combination: Half(40768,49,7,1) -> Half(21952,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce (CudnnConvolution) [TRT] Tactic: 0 Time: 0.686771 [TRT] Tactic: 1 Time: 0.630156 [TRT] Tactic: 2 Time: 0.74237 [TRT] Tactic: 4 skipped. Scratch requested: 862191616, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 51910656, available: 33554432 [TRT] Fastest Tactic: 1 Time: 0.630156 [TRT] --------------- Timing Runner: inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] Setting workspace to 51910656enables more tactics for profiling [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 1 [TRT] *************** Autotuning format combination: Half(20384,49:2,7,1) -> Half(21952,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] *************** Autotuning format combination: Half(20384,49:2,7,1) -> Half(10976,49:2,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.341016 [TRT] Tactic: 655359 Time: 0.373854 [TRT] Tactic: 786431 Time: 0.562553 [TRT] Tactic: 851967 Time: 0.451927 [TRT] Tactic: 1179647 Time: 0.294401 [TRT] Tactic: 1310719 Time: 0.545443 [TRT] Tactic: 1376255 Time: 0.343672 [TRT] Tactic: 1441791 Time: 0.400678 [TRT] Tactic: 1507327 Time: 0.413541 [TRT] Tactic: 1638399 Time: 0.501641 [TRT] Tactic: 1835007 Time: 0.468359 [TRT] Tactic: 1900543 Time: 0.570911 [TRT] Tactic: 2097151 Time: 0.398828 [TRT] Tactic: 2162687 Time: 0.38 [TRT] Tactic: 2293759 Time: 0.352917 [TRT] Tactic: 2359295 Time: 0.337734 [TRT] Tactic: 2686975 Time: 0.616641 [TRT] Tactic: 3080191 Time: 0.317291 [TRT] Tactic: 3342335 Time: 0.597135 [TRT] Tactic: 3407871 Time: 0.296198 [TRT] Tactic: 3538943 Time: 0.294688 [TRT] Tactic: 3670015 Time: 0.371901 [TRT] Tactic: 3932159 Time: 0.33487 [TRT] Tactic: 3997695 Time: 0.572891 [TRT] Tactic: 4063231 Time: 0.348958 [TRT] Tactic: 4194303 Time: 0.302917 [TRT] Tactic: 4259839 Time: 0.397501 [TRT] Tactic: 4325375 Time: 0.411171 [TRT] Tactic: 4521983 Time: 0.429219 [TRT] Tactic: 4587519 Time: 0.421198 [TRT] Tactic: 4653055 Time: 0.353151 [TRT] Tactic: 4915199 Time: 0.292031 [TRT] Tactic: 4980735 Time: 0.409401 [TRT] Tactic: 5177343 Time: 0.297604 [TRT] Tactic: 5242879 Time: 0.248177 [TRT] Tactic: 5373951 Time: 0.300287 [TRT] Tactic: 5439487 Time: 0.332344 [TRT] Tactic: 5570559 Time: 0.303932 [TRT] Tactic: 5636095 Time: 0.351302 [TRT] Tactic: 5701631 Time: 0.31211 [TRT] Tactic: 5767167 Time: 0.435859 [TRT] Tactic: 5832703 Time: 0.280235 [TRT] Tactic: 5898239 Time: 0.244193 [TRT] Tactic: 6029311 Time: 0.327396 [TRT] Tactic: 6225919 Time: 0.260208 [TRT] Tactic: 6291455 Time: 0.296588 [TRT] Tactic: 6422527 Time: 0.297422 [TRT] Tactic: 6750207 Time: 0.340807 [TRT] Tactic: 6815743 Time: 0.246667 [TRT] Tactic: 6946815 Time: 0.379401 [TRT] Tactic: 7012351 Time: 0.395833 [TRT] Tactic: 7077887 Time: 0.282344 [TRT] Tactic: 7143423 Time: 0.463437 [TRT] Tactic: 7208959 Time: 0.286068 [TRT] Tactic: 7340031 Time: 0.254584 [TRT] Tactic: 7405567 Time: 0.318021 [TRT] Tactic: 7536639 Time: 0.311172 [TRT] Tactic: 7602175 Time: 0.368906 [TRT] Tactic: 7733247 Time: 0.258854 [TRT] Tactic: 7798783 Time: 0.565703 [TRT] Tactic: 8191999 Time: 0.422109 [TRT] Tactic: 8257535 Time: 0.28836 [TRT] Tactic: 8323071 Time: 0.32638 [TRT] Tactic: 8650751 Time: 0.374583 [TRT] Tactic: 8716287 Time: 0.269974 [TRT] Tactic: 9109503 Time: 0.406823 [TRT] Tactic: 9568255 Time: 0.290729 [TRT] Tactic: 9895935 Time: 0.303958 [TRT] Tactic: 10223615 Time: 0.601041 [TRT] Tactic: 10354687 Time: 0.396667 [TRT] Tactic: 10551295 Time: 0.277656 [TRT] Tactic: 10747903 Time: 0.229661 [TRT] Tactic: 10944511 Time: 0.406302 [TRT] Fastest Tactic: 10747903 Time: 0.229661 [TRT] --------------- Timing Runner: inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce (CaskConvolution) [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_interior_nn_v1 Tactic: 3066127711859985668 [TRT] Tactic: 3066127711859985668 Time: 0.291172 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.344844 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.310989 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.265495 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] Tactic: 8163473458334948789 Time: 0.257812 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.287709 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.293307 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.265547 [TRT] inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_interior_nn_v1 Tactic: -1716393687483585322 [TRT] Tactic: -1716393687483585322 Time: 0.282266 [TRT] Fastest Tactic: 8163473458334948789 Time: 0.257812 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 10747903 [TRT] *************** Autotuning Reformat:Float(21952,49,7,1) -> Float(21952,1,3136,448) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0337495 [TRT] Tactic: 0 Time: 0.0370575 [TRT] Fastest Tactic: 1002 Time: 0.0337495 [TRT] *************** Autotuning Reformat:Float(21952,49,7,1) -> Half(21952,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.67974 [TRT] Tactic: 0 Time: 0.0265365 [TRT] Fastest Tactic: 0 Time: 0.0265365 [TRT] *************** Autotuning Reformat:Float(21952,49,7,1) -> Half(10976,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.042656 [TRT] Tactic: 0 Time: 0.021797 [TRT] Fastest Tactic: 0 Time: 0.021797 [TRT] *************** Autotuning Reformat:Float(21952,1,3136,448) -> Float(21952,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0439585 [TRT] Tactic: 0 Time: 0.0334115 [TRT] Fastest Tactic: 0 Time: 0.0334115 [TRT] *************** Autotuning Reformat:Float(21952,1,3136,448) -> Half(21952,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.033411 [TRT] Tactic: 0 Time: 0.0335675 [TRT] Fastest Tactic: 1002 Time: 0.033411 [TRT] *************** Autotuning Reformat:Float(21952,1,3136,448) -> Half(10976,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.038125 [TRT] Tactic: 0 Time: 0.0403385 [TRT] Fastest Tactic: 1002 Time: 0.038125 [TRT] *************** Autotuning Reformat:Half(21952,49,7,1) -> Float(21952,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.701744 [TRT] Tactic: 0 Time: 0.022083 [TRT] Fastest Tactic: 0 Time: 0.022083 [TRT] *************** Autotuning Reformat:Half(21952,49,7,1) -> Float(21952,1,3136,448) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0311455 [TRT] Tactic: 0 Time: 0.037917 [TRT] Fastest Tactic: 1002 Time: 0.0311455 [TRT] *************** Autotuning Reformat:Half(21952,49,7,1) -> Half(10976,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0466665 [TRT] Tactic: 0 Time: 0.021719 [TRT] Fastest Tactic: 0 Time: 0.021719 [TRT] *************** Autotuning Reformat:Half(10976,49:2,7,1) -> Float(21952,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.040521 [TRT] Tactic: 0 Time: 0.0209895 [TRT] Fastest Tactic: 0 Time: 0.0209895 [TRT] *************** Autotuning Reformat:Half(10976,49:2,7,1) -> Float(21952,1,3136,448) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0340885 [TRT] Tactic: 0 Time: 0.046823 [TRT] Fastest Tactic: 1002 Time: 0.0340885 [TRT] *************** Autotuning Reformat:Half(10976,49:2,7,1) -> Half(21952,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0571355 [TRT] Tactic: 0 Time: 0.021276 [TRT] Fastest Tactic: 0 Time: 0.021276 [TRT] *************** Autotuning Reformat:Float(21952,49,7,1) -> Float(21952,1,3136,448) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.017578 [TRT] Tactic: 0 Time: 0.016328 [TRT] Fastest Tactic: 0 Time: 0.016328 [TRT] *************** Autotuning Reformat:Float(21952,49,7,1) -> Half(21952,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.32375 [TRT] Tactic: 0 Time: 0.015833 [TRT] Fastest Tactic: 0 Time: 0.015833 [TRT] *************** Autotuning Reformat:Float(21952,49,7,1) -> Half(10976,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0212765 [TRT] Tactic: 0 Time: 0.018359 [TRT] Fastest Tactic: 0 Time: 0.018359 [TRT] *************** Autotuning Reformat:Float(21952,1,3136,448) -> Float(21952,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0209895 [TRT] Tactic: 0 Time: 0.016015 [TRT] Fastest Tactic: 0 Time: 0.016015 [TRT] *************** Autotuning Reformat:Float(21952,1,3136,448) -> Half(21952,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.016328 [TRT] Tactic: 0 Time: 0.0159375 [TRT] Fastest Tactic: 0 Time: 0.0159375 [TRT] *************** Autotuning Reformat:Float(21952,1,3136,448) -> Half(10976,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0208075 [TRT] Tactic: 0 Time: 0.0187495 [TRT] Fastest Tactic: 0 Time: 0.0187495 [TRT] *************** Autotuning Reformat:Half(21952,49,7,1) -> Float(21952,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.331744 [TRT] Tactic: 0 Time: 0.0108855 [TRT] Fastest Tactic: 0 Time: 0.0108855 [TRT] *************** Autotuning Reformat:Half(21952,49,7,1) -> Float(21952,1,3136,448) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.015625 [TRT] Tactic: 0 Time: 0.0174735 [TRT] Fastest Tactic: 1002 Time: 0.015625 [TRT] *************** Autotuning Reformat:Half(21952,49,7,1) -> Half(10976,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.022474 [TRT] Tactic: 0 Time: 0.010912 [TRT] Fastest Tactic: 0 Time: 0.010912 [TRT] *************** Autotuning Reformat:Half(10976,49:2,7,1) -> Float(21952,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.02125 [TRT] Tactic: 0 Time: 0.010781 [TRT] Fastest Tactic: 0 Time: 0.010781 [TRT] *************** Autotuning Reformat:Half(10976,49:2,7,1) -> Float(21952,1,3136,448) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0157815 [TRT] Tactic: 0 Time: 0.0188025 [TRT] Fastest Tactic: 1002 Time: 0.0157815 [TRT] *************** Autotuning Reformat:Half(10976,49:2,7,1) -> Half(21952,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.028489 [TRT] Tactic: 0 Time: 0.016094 [TRT] Fastest Tactic: 0 Time: 0.016094 [TRT] *************** Autotuning format combination: Float(21952,49,7,1) -> Float(40768,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/3x3 + inception_5a/relu_3x3 (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5a/3x3 + inception_5a/relu_3x3 (FusedConvActConvolution) [TRT] Tactic: 524287 Time: 0.384427 [TRT] Tactic: 720895 Time: 0.655391 [TRT] Tactic: 983039 Time: 0.328959 [TRT] Tactic: 1048575 Time: 0.496406 [TRT] Tactic: 1703935 Time: 0.476693 [TRT] Tactic: 1769471 Time: 0.437604 [TRT] Tactic: 1966079 Time: 0.46901 [TRT] Tactic: 2031615 Time: 0.680208 [TRT] Tactic: 2228223 Time: 0.676718 [TRT] Tactic: 2424831 Time: 0.681094 [TRT] Tactic: 2621439 Time: 0.570312 [TRT] Tactic: 2752511 Time: 0.492526 [TRT] Tactic: 2818047 Time: 0.700938 [TRT] Tactic: 2883583 Time: 0.577187 [TRT] Tactic: 3014655 Time: 0.367422 [TRT] Tactic: 3145727 Time: 0.450703 [TRT] Tactic: 3473407 Time: 0.587709 [TRT] Tactic: 3604479 Time: 0.360807 [TRT] Tactic: 3735551 Time: 0.726094 [TRT] Tactic: 4390911 Time: 0.527265 [TRT] Tactic: 5046271 Time: 0.404714 [TRT] Tactic: 5963775 Time: 0.415781 [TRT] Tactic: 6160383 Time: 0.407812 [TRT] Tactic: 6488063 Time: 0.372422 [TRT] Tactic: 6881279 Time: 0.44651 [TRT] Tactic: 7274495 Time: 0.497994 [TRT] Tactic: 7864319 Time: 0.594505 [TRT] Tactic: 7995391 Time: 0.439792 [TRT] Tactic: 8585215 Time: 0.4775 [TRT] Tactic: 8847359 Time: 0.351667 [TRT] Tactic: 8978431 Time: 0.424401 [TRT] Tactic: 9043967 Time: 0.33362 [TRT] Tactic: 9175039 Time: 0.363932 [TRT] Tactic: 9502719 Time: 0.542943 [TRT] Tactic: 9830399 Time: 0.448724 [TRT] Tactic: 9961471 Time: 0.390833 [TRT] Tactic: 10027007 Time: 0.464193 [TRT] Tactic: 10092543 Time: 0.52599 [TRT] Tactic: 10289151 Time: 0.468489 [TRT] Tactic: 10485759 Time: 0.455208 [TRT] Tactic: 10682367 Time: 0.57263 [TRT] Tactic: 10813439 Time: 0.418177 [TRT] Fastest Tactic: 983039 Time: 0.328959 [TRT] --------------- Timing Runner: inception_5a/3x3 + inception_5a/relu_3x3 (CudnnConvolution) [TRT] Tactic: 0 Time: 0.843125 [TRT] Tactic: 1 Time: 0.938385 [TRT] Tactic: 2 Time: 0.903021 [TRT] Tactic: 4 skipped. Scratch requested: 120176640, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 224911360, available: 33554432 [TRT] Tactic: 6 Time: 0.853099 [TRT] Fastest Tactic: 0 Time: 0.843125 [TRT] --------------- Timing Runner: inception_5a/3x3 + inception_5a/relu_3x3 (CaskConvolution) [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 1.01789 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v0 Tactic: 1754984623894446479 [TRT] Tactic: 1754984623894446479 Time: 1.202 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v0 Tactic: 3611739942397549984 [TRT] Tactic: 3611739942397549984 Time: 1.01779 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148n_nt_v1 Tactic: 3827454225649558724 [TRT] Tactic: 3827454225649558724 Time: 0.582136 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v1 Tactic: 4337000649858996379 [TRT] Tactic: 4337000649858996379 Time: 0.976615 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 1.0232 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.874688 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.988724 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 5921334924264294896 [TRT] Tactic: 5921334924264294896 Time: 0.46776 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.95211 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.990104 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148t_nt_v1 Tactic: 7852627285308570038 [TRT] Tactic: 7852627285308570038 Time: 0.568411 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v1 Tactic: -9137461792520977713 [TRT] Tactic: -9137461792520977713 Time: 1.04159 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148n_nt_v0 Tactic: -8776506421218919509 [TRT] Tactic: -8776506421218919509 Time: 0.540209 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 1.02391 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v0 Tactic: -8133971918129952780 [TRT] Tactic: -8133971918129952780 Time: 1.12651 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v1 Tactic: -6092040395344634144 [TRT] Tactic: -6092040395344634144 Time: 1.18612 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 1.19992 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.945416 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148t_nt_v0 Tactic: -2318106587342035239 [TRT] Tactic: -2318106587342035239 Time: 0.542943 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_mobile_relu_tile148t_nt_v0 Tactic: -1343271414618805657 [TRT] Tactic: -1343271414618805657 Time: 0.451093 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 1.10891 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 1.0407 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.965417 [TRT] Fastest Tactic: -1343271414618805657 Time: 0.451093 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 983039 [TRT] *************** Autotuning format combination: Float(21952,1,3136,448) -> Float(40768,1,5824,832) *************** [TRT] --------------- Timing Runner: inception_5a/3x3 + inception_5a/relu_3x3 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5a/3x3 + inception_5a/relu_3x3 (CaskConvolution) [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.94961 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.80586 [TRT] Fastest Tactic: -7394439838318485025 Time: 0.80586 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(21952,49,7,1) -> Half(40768,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/3x3 + inception_5a/relu_3x3 (CudnnConvolution) [TRT] Tactic: 0 Time: 1.22599 [TRT] Tactic: 1 Time: 1.2474 [TRT] Tactic: 2 Time: 0.9775 [TRT] Tactic: 4 skipped. Scratch requested: 120176640, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 224911360, available: 33554432 [TRT] Tactic: 6 Time: 1.89013 [TRT] Fastest Tactic: 2 Time: 0.9775 [TRT] --------------- Timing Runner: inception_5a/3x3 + inception_5a/relu_3x3 (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 2 [TRT] *************** Autotuning format combination: Half(10976,49:2,7,1) -> Half(20384,49:2,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/3x3 + inception_5a/relu_3x3 (FusedConvActConvolution) [TRT] Tactic: 524287 Time: 0.250365 [TRT] Tactic: 720895 Time: 0.464895 [TRT] Tactic: 983039 Time: 0.224401 [TRT] Tactic: 1048575 Time: 0.331927 [TRT] Tactic: 1703935 Time: 0.327631 [TRT] Tactic: 1769471 Time: 1.59562 [TRT] Tactic: 1966079 Time: 0.288698 [TRT] Tactic: 2031615 Time: 0.395885 [TRT] Tactic: 2228223 Time: 0.421979 [TRT] Tactic: 2424831 Time: 0.553203 [TRT] Tactic: 2621439 Time: 0.350026 [TRT] Tactic: 2752511 Time: 0.288359 [TRT] Tactic: 2818047 Time: 0.426615 [TRT] Tactic: 2883583 Time: 0.415 [TRT] Tactic: 3014655 Time: 0.239166 [TRT] Tactic: 3145727 Time: 0.287135 [TRT] Tactic: 3473407 Time: 0.43724 [TRT] Tactic: 3604479 Time: 0.235339 [TRT] Tactic: 3735551 Time: 0.414193 [TRT] Tactic: 4390911 Time: 0.347735 [TRT] Tactic: 5046271 Time: 0.245703 [TRT] Tactic: 5963775 Time: 0.272838 [TRT] Tactic: 6160383 Time: 0.262865 [TRT] Tactic: 6488063 Time: 0.229818 [TRT] Tactic: 6881279 Time: 0.274635 [TRT] Tactic: 7274495 Time: 0.306146 [TRT] Tactic: 7864319 Time: 0.379036 [TRT] Tactic: 7995391 Time: 0.271042 [TRT] Tactic: 8585215 Time: 0.312344 [TRT] Tactic: 8847359 Time: 0.214454 [TRT] Tactic: 8978431 Time: 0.255964 [TRT] Tactic: 9043967 Time: 0.210443 [TRT] Tactic: 9175039 Time: 0.236458 [TRT] Tactic: 9502719 Time: 0.338308 [TRT] Tactic: 9830399 Time: 0.28888 [TRT] Tactic: 9961471 Time: 0.24073 [TRT] Tactic: 10027007 Time: 0.29026 [TRT] Tactic: 10092543 Time: 0.346901 [TRT] Tactic: 10289151 Time: 0.29987 [TRT] Tactic: 10485759 Time: 0.27836 [TRT] Tactic: 10682367 Time: 0.344505 [TRT] Tactic: 10813439 Time: 0.26401 [TRT] Fastest Tactic: 9043967 Time: 0.210443 [TRT] --------------- Timing Runner: inception_5a/3x3 + inception_5a/relu_3x3 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5a/3x3 + inception_5a/relu_3x3 (CaskConvolution) [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.589818 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_large_nn_v1 Tactic: 3650389455493082349 [TRT] Tactic: 3650389455493082349 Time: 0.599844 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_winograd_fp16x2_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 4772821744921268633 [TRT] Tactic: 4772821744921268633 Time: 0.254297 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.506172 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.470209 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_large_nn_v1 Tactic: -6490690591794140522 [TRT] Tactic: -6490690591794140522 Time: 0.481953 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_large_nn_v1 Tactic: -4686027666808657977 [TRT] Tactic: -4686027666808657977 Time: 0.520678 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.488437 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.516797 [TRT] inception_5a/3x3 + inception_5a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.422083 [TRT] Fastest Tactic: 4772821744921268633 Time: 0.254297 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 9043967 [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Float(40768,1,5824,832) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.033021 [TRT] Tactic: 0 Time: 0.0361975 [TRT] Fastest Tactic: 1002 Time: 0.033021 [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Half(40768,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.786276 [TRT] Tactic: 0 Time: 0.033099 [TRT] Fastest Tactic: 0 Time: 0.033099 [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Half(20384,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0424995 [TRT] Tactic: 0 Time: 0.021484 [TRT] Fastest Tactic: 0 Time: 0.021484 [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Float(40768,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.041953 [TRT] Tactic: 0 Time: 0.033464 [TRT] Fastest Tactic: 0 Time: 0.033464 [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Half(40768,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.033385 [TRT] Tactic: 0 Time: 0.033125 [TRT] Fastest Tactic: 0 Time: 0.033125 [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Half(20384,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0385155 [TRT] Tactic: 0 Time: 0.03862 [TRT] Fastest Tactic: 1002 Time: 0.0385155 [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Float(40768,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.804245 [TRT] Tactic: 0 Time: 0.033047 [TRT] Fastest Tactic: 0 Time: 0.033047 [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Float(40768,1,5824,832) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.030807 [TRT] Tactic: 0 Time: 0.036198 [TRT] Fastest Tactic: 1002 Time: 0.030807 [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Half(20384,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0465365 [TRT] Tactic: 0 Time: 0.0224735 [TRT] Fastest Tactic: 0 Time: 0.0224735 [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Float(40768,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0415885 [TRT] Tactic: 0 Time: 0.0383335 [TRT] Fastest Tactic: 0 Time: 0.0383335 [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Float(40768,1,5824,832) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.030755 [TRT] Tactic: 0 Time: 0.041172 [TRT] Fastest Tactic: 1002 Time: 0.030755 [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Half(40768,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0524215 [TRT] Tactic: 0 Time: 0.0358065 [TRT] Fastest Tactic: 0 Time: 0.0358065 [TRT] *************** Autotuning Reformat:Float(21952,49,7,1) -> Float(21952,1,3136,448) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0182815 [TRT] Tactic: 0 Time: 0.0090885 [TRT] Fastest Tactic: 0 Time: 0.0090885 [TRT] *************** Autotuning Reformat:Float(21952,49,7,1) -> Half(21952,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.084375 [TRT] Tactic: 0 Time: 0.008073 [TRT] Fastest Tactic: 0 Time: 0.008073 [TRT] *************** Autotuning Reformat:Float(21952,49,7,1) -> Half(10976,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0119005 [TRT] Tactic: 0 Time: 0.009063 [TRT] Fastest Tactic: 0 Time: 0.009063 [TRT] *************** Autotuning Reformat:Float(21952,1,3136,448) -> Float(21952,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.01875 [TRT] Tactic: 0 Time: 0.0079165 [TRT] Fastest Tactic: 0 Time: 0.0079165 [TRT] *************** Autotuning Reformat:Float(21952,1,3136,448) -> Half(21952,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0116925 [TRT] Tactic: 0 Time: 0.0078125 [TRT] Fastest Tactic: 0 Time: 0.0078125 [TRT] *************** Autotuning Reformat:Float(21952,1,3136,448) -> Half(10976,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.011589 [TRT] Tactic: 0 Time: 0.008985 [TRT] Fastest Tactic: 0 Time: 0.008985 [TRT] *************** Autotuning Reformat:Half(21952,49,7,1) -> Float(21952,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.091927 [TRT] Tactic: 0 Time: 0.007031 [TRT] Fastest Tactic: 0 Time: 0.007031 [TRT] *************** Autotuning Reformat:Half(21952,49,7,1) -> Float(21952,1,3136,448) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0126825 [TRT] Tactic: 0 Time: 0.009792 [TRT] Fastest Tactic: 0 Time: 0.009792 [TRT] *************** Autotuning Reformat:Half(21952,49,7,1) -> Half(10976,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.015573 [TRT] Tactic: 0 Time: 0.006927 [TRT] Fastest Tactic: 0 Time: 0.006927 [TRT] *************** Autotuning Reformat:Half(10976,49:2,7,1) -> Float(21952,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.014375 [TRT] Tactic: 0 Time: 0.006979 [TRT] Fastest Tactic: 0 Time: 0.006979 [TRT] *************** Autotuning Reformat:Half(10976,49:2,7,1) -> Float(21952,1,3136,448) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0127605 [TRT] Tactic: 0 Time: 0.009896 [TRT] Fastest Tactic: 0 Time: 0.009896 [TRT] *************** Autotuning Reformat:Half(10976,49:2,7,1) -> Half(21952,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.018437 [TRT] Tactic: 0 Time: 0.009792 [TRT] Fastest Tactic: 0 Time: 0.009792 [TRT] *************** Autotuning format combination: Float(21952,49,7,1) -> Float(40768,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/5x5 + inception_5a/relu_5x5 (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5a/5x5 + inception_5a/relu_5x5 (FusedConvActConvolution) [TRT] Tactic: 393215 Time: 0.147839 [TRT] Tactic: 917503 Time: 0.2075 [TRT] Tactic: 1114111 Time: 0.175104 [TRT] Tactic: 1245183 Time: 0.206198 [TRT] Tactic: 1572863 Time: 0.218464 [TRT] Tactic: 2490367 Time: 0.189037 [TRT] Tactic: 2555903 Time: 0.229687 [TRT] Tactic: 2949119 Time: 0.180339 [TRT] Tactic: 3211263 Time: 0.182734 [TRT] Tactic: 3801087 Time: 0.192448 [TRT] Tactic: 3866623 Time: 0.166953 [TRT] Tactic: 4128767 Time: 0.175469 [TRT] Tactic: 4456447 Time: 0.19388 [TRT] Tactic: 4718591 Time: 0.167187 [TRT] Tactic: 4784127 Time: 0.278672 [TRT] Tactic: 4849663 Time: 0.16862 [TRT] Tactic: 5111807 Time: 0.171562 [TRT] Tactic: 5308415 Time: 0.201822 [TRT] Tactic: 5505023 Time: 0.181745 [TRT] Tactic: 6094847 Time: 0.193802 [TRT] Tactic: 6356991 Time: 0.225652 [TRT] Tactic: 6553599 Time: 0.145 [TRT] Tactic: 6619135 Time: 0.166667 [TRT] Tactic: 6684671 Time: 0.154974 [TRT] Tactic: 7471103 Time: 0.173672 [TRT] Tactic: 7667711 Time: 0.165339 [TRT] Tactic: 7929855 Time: 0.149271 [TRT] Tactic: 8060927 Time: 0.140625 [TRT] Tactic: 8126463 Time: 0.189452 [TRT] Tactic: 8388607 Time: 0.140469 [TRT] Tactic: 8519679 Time: 0.193985 [TRT] Tactic: 8781823 Time: 0.178854 [TRT] Tactic: 8912895 Time: 0.189323 [TRT] Tactic: 9240575 Time: 0.177604 [TRT] Tactic: 9306111 Time: 0.211432 [TRT] Tactic: 9371647 Time: 0.14862 [TRT] Tactic: 9437183 Time: 0.178567 [TRT] Tactic: 9633791 Time: 0.169089 [TRT] Tactic: 9699327 Time: 0.138359 [TRT] Tactic: 9764863 Time: 0.185911 [TRT] Tactic: 10158079 Time: 0.143724 [TRT] Tactic: 10420223 Time: 0.155651 [TRT] Tactic: 10616831 Time: 0.192787 [TRT] Tactic: 10878975 Time: 0.200078 [TRT] Fastest Tactic: 9699327 Time: 0.138359 [TRT] --------------- Timing Runner: inception_5a/5x5 + inception_5a/relu_5x5 (CudnnConvolution) [TRT] Tactic: 0 Time: 0.364922 [TRT] Tactic: 1 Time: 0.363697 [TRT] Tactic: 2 Time: 0.406849 [TRT] Tactic: 4 Time: 2.24953 [TRT] Tactic: 5 Time: 8.01484 [TRT] Fastest Tactic: 1 Time: 0.363697 [TRT] --------------- Timing Runner: inception_5a/5x5 + inception_5a/relu_5x5 (CaskConvolution) [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.379349 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v0 Tactic: 1754984623894446479 [TRT] Tactic: 1754984623894446479 Time: 0.396042 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v0 Tactic: 3611739942397549984 [TRT] Tactic: 3611739942397549984 Time: 0.26237 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v1 Tactic: 4337000649858996379 [TRT] Tactic: 4337000649858996379 Time: 0.303281 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.2575 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.232005 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.238308 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.29823 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.278984 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v1 Tactic: -9137461792520977713 [TRT] Tactic: -9137461792520977713 Time: 0.279114 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.27164 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v0 Tactic: -8133971918129952780 [TRT] Tactic: -8133971918129952780 Time: 0.341354 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v1 Tactic: -6092040395344634144 [TRT] Tactic: -6092040395344634144 Time: 0.392969 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.391589 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.296588 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.339661 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.278646 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.227578 [TRT] Fastest Tactic: -410470605513481746 Time: 0.227578 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 9699327 [TRT] *************** Autotuning format combination: Float(21952,1,3136,448) -> Float(40768,1,5824,832) *************** [TRT] --------------- Timing Runner: inception_5a/5x5 + inception_5a/relu_5x5 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5a/5x5 + inception_5a/relu_5x5 (CaskConvolution) [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.261172 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.205052 [TRT] Fastest Tactic: -7394439838318485025 Time: 0.205052 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(21952,49,7,1) -> Half(40768,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/5x5 + inception_5a/relu_5x5 (CudnnConvolution) [TRT] Tactic: 0 Time: 0.369062 [TRT] Tactic: 1 Time: 0.683177 [TRT] Tactic: 2 Time: 0.425209 [TRT] Tactic: 4 Time: 2.28133 [TRT] Tactic: 5 Time: 7.78609 [TRT] Fastest Tactic: 0 Time: 0.369062 [TRT] --------------- Timing Runner: inception_5a/5x5 + inception_5a/relu_5x5 (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 0 [TRT] *************** Autotuning format combination: Half(10976,49:2,7,1) -> Half(20384,49:2,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/5x5 + inception_5a/relu_5x5 (FusedConvActConvolution) [TRT] Tactic: 393215 Time: 0.0733595 [TRT] Tactic: 917503 Time: 0.111328 [TRT] Tactic: 1114111 Time: 0.082682 [TRT] Tactic: 1245183 Time: 0.098177 [TRT] Tactic: 1572863 Time: 0.111927 [TRT] Tactic: 2490367 Time: 0.110443 [TRT] Tactic: 2555903 Time: 0.137969 [TRT] Tactic: 2949119 Time: 0.0902345 [TRT] Tactic: 3211263 Time: 0.089636 [TRT] Tactic: 3801087 Time: 0.0970835 [TRT] Tactic: 3866623 Time: 0.0854425 [TRT] Tactic: 4128767 Time: 0.0888805 [TRT] Tactic: 4456447 Time: 0.103542 [TRT] Tactic: 4718591 Time: 0.081432 [TRT] Tactic: 4784127 Time: 0.136485 [TRT] Tactic: 4849663 Time: 0.085338 [TRT] Tactic: 5111807 Time: 0.0832035 [TRT] Tactic: 5308415 Time: 0.098099 [TRT] Tactic: 5505023 Time: 0.0861725 [TRT] Tactic: 6094847 Time: 0.0993225 [TRT] Tactic: 6356991 Time: 0.135911 [TRT] Tactic: 6553599 Time: 0.0732555 [TRT] Tactic: 6619135 Time: 0.080834 [TRT] Tactic: 6684671 Time: 0.072839 [TRT] Tactic: 7471103 Time: 0.089297 [TRT] Tactic: 7667711 Time: 0.081458 [TRT] Tactic: 7929855 Time: 0.068984 [TRT] Tactic: 8060927 Time: 0.0747135 [TRT] Tactic: 8126463 Time: 0.0829945 [TRT] Tactic: 8388607 Time: 0.0708075 [TRT] Tactic: 8519679 Time: 0.086797 [TRT] Tactic: 8781823 Time: 0.0851825 [TRT] Tactic: 8912895 Time: 0.0952865 [TRT] Tactic: 9240575 Time: 0.078776 [TRT] Tactic: 9306111 Time: 0.097526 [TRT] Tactic: 9371647 Time: 0.070495 [TRT] Tactic: 9437183 Time: 0.0897915 [TRT] Tactic: 9633791 Time: 0.083828 [TRT] Tactic: 9699327 Time: 0.0709115 [TRT] Tactic: 9764863 Time: 0.0984375 [TRT] Tactic: 10158079 Time: 0.0738805 [TRT] Tactic: 10420223 Time: 0.0809635 [TRT] Tactic: 10616831 Time: 0.0986455 [TRT] Tactic: 10878975 Time: 0.104401 [TRT] Fastest Tactic: 7929855 Time: 0.068984 [TRT] --------------- Timing Runner: inception_5a/5x5 + inception_5a/relu_5x5 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5a/5x5 + inception_5a/relu_5x5 (CaskConvolution) [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.184375 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_large_nn_v1 Tactic: 3650389455493082349 [TRT] Tactic: 3650389455493082349 Time: 0.190078 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.133281 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.140573 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_large_nn_v1 Tactic: -6490690591794140522 [TRT] Tactic: -6490690591794140522 Time: 0.144219 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_large_nn_v1 Tactic: -4686027666808657977 [TRT] Tactic: -4686027666808657977 Time: 0.13026 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.106823 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.120573 [TRT] inception_5a/5x5 + inception_5a/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.114115 [TRT] Fastest Tactic: -4212163711445252890 Time: 0.106823 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 7929855 [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Float(40768,1,5824,832) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.017318 [TRT] Tactic: 0 Time: 0.0172395 [TRT] Fastest Tactic: 0 Time: 0.0172395 [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Half(40768,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.317057 [TRT] Tactic: 0 Time: 0.0167185 [TRT] Fastest Tactic: 0 Time: 0.0167185 [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Half(20384,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0227085 [TRT] Tactic: 0 Time: 0.011562 [TRT] Fastest Tactic: 0 Time: 0.011562 [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Float(40768,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0224995 [TRT] Tactic: 0 Time: 0.017084 [TRT] Fastest Tactic: 0 Time: 0.017084 [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Half(40768,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0172135 [TRT] Tactic: 0 Time: 0.016927 [TRT] Fastest Tactic: 0 Time: 0.016927 [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Half(20384,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.021406 [TRT] Tactic: 0 Time: 0.0196615 [TRT] Fastest Tactic: 0 Time: 0.0196615 [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Float(40768,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.326016 [TRT] Tactic: 0 Time: 0.016823 [TRT] Fastest Tactic: 0 Time: 0.016823 [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Float(40768,1,5824,832) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.017136 [TRT] Tactic: 0 Time: 0.017318 [TRT] Fastest Tactic: 1002 Time: 0.017136 [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Half(20384,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.026406 [TRT] Tactic: 0 Time: 0.0115105 [TRT] Fastest Tactic: 0 Time: 0.0115105 [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Float(40768,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0222915 [TRT] Tactic: 0 Time: 0.018646 [TRT] Fastest Tactic: 0 Time: 0.018646 [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Float(40768,1,5824,832) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.017031 [TRT] Tactic: 0 Time: 0.019974 [TRT] Fastest Tactic: 1002 Time: 0.017031 [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Half(40768,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.027656 [TRT] Tactic: 0 Time: 0.016823 [TRT] Fastest Tactic: 0 Time: 0.016823 [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Half(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Half(20384,49:2,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Float(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Half(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Half(20384,49:2,7,1) *************** [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Float(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Half(20384,49:2,7,1) *************** [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Float(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Half(40768,49,7,1) *************** [TRT] *************** Autotuning format combination: Float(40768,49,7,1) -> Float(40768,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/pool (TiledPooling) [TRT] Tactic: 2752769 Time: 0.119817 [TRT] Tactic: 2818305 Time: 0.113724 [TRT] Tactic: 2883841 Time: 0.30099 [TRT] Tactic: 2949377 Time: 1.05987 [TRT] Tactic: 3014913 Time: 0.782865 [TRT] Tactic: 3080449 Time: 0.43927 [TRT] Tactic: 3145985 Time: 0.358124 [TRT] Tactic: 3211521 Time: 0.0752865 [TRT] Tactic: 3277057 Time: 0.0743745 [TRT] Tactic: 3342593 Time: 0.17198 [TRT] Tactic: 3408129 Time: 0.604219 [TRT] Tactic: 3473665 Time: 0.448671 [TRT] Tactic: 3539201 Time: 0.264167 [TRT] Tactic: 3604737 Time: 0.218619 [TRT] Tactic: 3670273 Time: 0.063985 [TRT] Tactic: 3735809 Time: 0.0614325 [TRT] Tactic: 3801345 Time: 0.137396 [TRT] Tactic: 3866881 Time: 0.472475 [TRT] Tactic: 3932417 Time: 0.346927 [TRT] Tactic: 3997953 Time: 0.211484 [TRT] Tactic: 4063489 Time: 0.17849 [TRT] Tactic: 4129025 Time: 0.0564845 [TRT] Tactic: 4194561 Time: 0.055964 [TRT] Tactic: 4260097 Time: 0.119818 [TRT] Tactic: 4325633 Time: 0.401536 [TRT] Tactic: 4391169 Time: 0.295 [TRT] Tactic: 4456705 Time: 0.181823 [TRT] Tactic: 4522241 Time: 0.157891 [TRT] Tactic: 4587777 Time: 0.0544535 [TRT] Tactic: 4653313 Time: 0.0535415 [TRT] Tactic: 4718849 Time: 0.113646 [TRT] Tactic: 4784385 Time: 0.371771 [TRT] Tactic: 4849921 Time: 0.27612 [TRT] Tactic: 4915457 Time: 0.174219 [TRT] Tactic: 4980993 Time: 0.147579 [TRT] Tactic: 5046529 Time: 0.0537505 [TRT] Tactic: 5112065 Time: 0.0521355 [TRT] Tactic: 5177601 Time: 0.106797 [TRT] Tactic: 5243137 Time: 0.344974 [TRT] Tactic: 5308673 Time: 0.256094 [TRT] Tactic: 5374209 Time: 0.161953 [TRT] Tactic: 5439745 Time: 0.139505 [TRT] Tactic: 6553857 Time: 0.17427 [TRT] Tactic: 6750465 Time: 0.116692 [TRT] Fastest Tactic: 5112065 Time: 0.0521355 [TRT] --------------- Timing Runner: inception_5a/pool (CudnnPooling) [TRT] Tactic: -1 Time: 0.170026 [TRT] Fastest Tactic: -1 Time: 0.170026 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: TiledPooling Tactic: 5112065 [TRT] *************** Autotuning format combination: Half(40768,49,7,1) -> Half(40768,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/pool (TiledPooling) [TRT] TiledPooling has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5a/pool (CudnnPooling) [TRT] Tactic: -1 Time: 0.177213 [TRT] Fastest Tactic: -1 Time: 0.177213 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnPooling Tactic: -1 [TRT] *************** Autotuning format combination: Half(20384,49:2,7,1) -> Half(20384,49:2,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/pool (TiledPooling) [TRT] Tactic: 2752769 Time: 0.067839 [TRT] Tactic: 2818305 Time: 0.0677085 [TRT] Tactic: 2883841 Time: 0.170391 [TRT] Tactic: 2949377 Time: 0.573619 [TRT] Tactic: 3014913 Time: 0.439297 [TRT] Tactic: 3080449 Time: 0.250834 [TRT] Tactic: 3145985 Time: 0.201849 [TRT] Tactic: 3211521 Time: 0.0481775 [TRT] Tactic: 3277057 Time: 0.046328 [TRT] Tactic: 3342593 Time: 0.121068 [TRT] Tactic: 3408129 Time: 0.379844 [TRT] Tactic: 3473665 Time: 0.286198 [TRT] Tactic: 3539201 Time: 0.167552 [TRT] Tactic: 3604737 Time: 0.142058 [TRT] Tactic: 3670273 Time: 0.0416145 [TRT] Tactic: 3735809 Time: 0.0412495 [TRT] Tactic: 3801345 Time: 0.106798 [TRT] Tactic: 3866881 Time: 0.323541 [TRT] Tactic: 3932417 Time: 0.244271 [TRT] Tactic: 3997953 Time: 0.145078 [TRT] Tactic: 4063489 Time: 0.124766 [TRT] Tactic: 4129025 Time: 0.0386195 [TRT] Tactic: 4194561 Time: 0.038724 [TRT] Tactic: 4260097 Time: 0.09987 [TRT] Tactic: 4325633 Time: 0.285886 [TRT] Tactic: 4391169 Time: 0.222708 [TRT] Tactic: 4456705 Time: 0.133255 [TRT] Tactic: 4522241 Time: 0.112683 [TRT] Tactic: 4587777 Time: 0.03862 [TRT] Tactic: 4653313 Time: 0.038646 [TRT] Tactic: 4718849 Time: 0.096302 [TRT] Tactic: 4784385 Time: 0.273776 [TRT] Tactic: 4849921 Time: 0.210834 [TRT] Tactic: 4915457 Time: 0.127292 [TRT] Tactic: 4980993 Time: 0.110078 [TRT] Tactic: 5046529 Time: 0.0376305 [TRT] Tactic: 5112065 Time: 0.036276 [TRT] Tactic: 5177601 Time: 0.0945575 [TRT] Tactic: 5243137 Time: 0.263359 [TRT] Tactic: 5308673 Time: 0.201927 [TRT] Tactic: 5374209 Time: 0.123125 [TRT] Tactic: 5439745 Time: 0.107318 [TRT] Tactic: 6553857 Time: 0.109818 [TRT] Tactic: 6750465 Time: 0.078984 [TRT] Fastest Tactic: 5112065 Time: 0.036276 [TRT] --------------- Timing Runner: inception_5a/pool (CudaPooling) [TRT] Tactic: -3 Time: 0.099479 [TRT] Fastest Tactic: -3 Time: 0.099479 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: TiledPooling Tactic: 5112065 [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Float(40768,1,5824,832) *************** [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Half(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Half(20384,49:2,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Float(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Half(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Half(20384,49:2,7,1) *************** [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Float(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Float(40768,1,5824,832) *************** [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Half(20384,49:2,7,1) *************** [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Float(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Float(40768,1,5824,832) *************** [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Half(40768,49,7,1) *************** [TRT] *************** Autotuning format combination: Float(40768,49,7,1) -> Float(40768,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/pool_proj + inception_5a/relu_pool_proj (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5a/pool_proj + inception_5a/relu_pool_proj (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.341589 [TRT] Tactic: 655359 Time: 0.199349 [TRT] Tactic: 786431 Time: 0.345833 [TRT] Tactic: 851967 Time: 0.220756 [TRT] Tactic: 1179647 Time: 0.212084 [TRT] Tactic: 1310719 Time: 0.533802 [TRT] Tactic: 1376255 Time: 0.375313 [TRT] Tactic: 1441791 Time: 0.269401 [TRT] Tactic: 1507327 Time: 0.229714 [TRT] Tactic: 1638399 Time: 0.332214 [TRT] Tactic: 1835007 Time: 0.34336 [TRT] Tactic: 1900543 Time: 0.428489 [TRT] Tactic: 2097151 Time: 0.213021 [TRT] Tactic: 2162687 Time: 0.416563 [TRT] Tactic: 2293759 Time: 0.393933 [TRT] Tactic: 2359295 Time: 0.249427 [TRT] Tactic: 2686975 Time: 0.353568 [TRT] Tactic: 3080191 Time: 0.150599 [TRT] Tactic: 3342335 Time: 0.420807 [TRT] Tactic: 3407871 Time: 0.225938 [TRT] Tactic: 3538943 Time: 0.231406 [TRT] Tactic: 3670015 Time: 0.348074 [TRT] Tactic: 3932159 Time: 0.34763 [TRT] Tactic: 3997695 Time: 0.364114 [TRT] Tactic: 4063231 Time: 0.22625 [TRT] Tactic: 4194303 Time: 0.192995 [TRT] Tactic: 4259839 Time: 0.229896 [TRT] Tactic: 4325375 Time: 0.323308 [TRT] Tactic: 4521983 Time: 0.390051 [TRT] Tactic: 4587519 Time: 0.292109 [TRT] Tactic: 4653055 Time: 0.238229 [TRT] Tactic: 4915199 Time: 0.200547 [TRT] Tactic: 4980735 Time: 0.303255 [TRT] Tactic: 5177343 Time: 0.211329 [TRT] Tactic: 5242879 Time: 0.214245 [TRT] Tactic: 5373951 Time: 0.214062 [TRT] Tactic: 5439487 Time: 0.202657 [TRT] Tactic: 5570559 Time: 0.150937 [TRT] Tactic: 5636095 Time: 0.227109 [TRT] Tactic: 5701631 Time: 0.386328 [TRT] Tactic: 5767167 Time: 0.271901 [TRT] Tactic: 5832703 Time: 0.224817 [TRT] Tactic: 5898239 Time: 0.166588 [TRT] Tactic: 6029311 Time: 0.364532 [TRT] Tactic: 6225919 Time: 0.16026 [TRT] Tactic: 6291455 Time: 0.212995 [TRT] Tactic: 6422527 Time: 0.192396 [TRT] Tactic: 6750207 Time: 0.202787 [TRT] Tactic: 6815743 Time: 0.21401 [TRT] Tactic: 6946815 Time: 0.342708 [TRT] Tactic: 7012351 Time: 0.212968 [TRT] Tactic: 7077887 Time: 0.227499 [TRT] Tactic: 7143423 Time: 0.288541 [TRT] Tactic: 7208959 Time: 0.224245 [TRT] Tactic: 7340031 Time: 0.170442 [TRT] Tactic: 7405567 Time: 0.225729 [TRT] Tactic: 7536639 Time: 0.22276 [TRT] Tactic: 7602175 Time: 0.338177 [TRT] Tactic: 7733247 Time: 0.149531 [TRT] Tactic: 7798783 Time: 0.347032 [TRT] Tactic: 8191999 Time: 0.358724 [TRT] Tactic: 8257535 Time: 0.202187 [TRT] Tactic: 8323071 Time: 0.204062 [TRT] Tactic: 8650751 Time: 0.338568 [TRT] Tactic: 8716287 Time: 0.161302 [TRT] Tactic: 9109503 Time: 0.213593 [TRT] Tactic: 9568255 Time: 0.199766 [TRT] Tactic: 9895935 Time: 0.192708 [TRT] Tactic: 10223615 Time: 0.357136 [TRT] Tactic: 10354687 Time: 0.241302 [TRT] Tactic: 10551295 Time: 0.242579 [TRT] Tactic: 10747903 Time: 0.14987 [TRT] Tactic: 10944511 Time: 0.304219 [TRT] Fastest Tactic: 7733247 Time: 0.149531 [TRT] --------------- Timing Runner: inception_5a/pool_proj + inception_5a/relu_pool_proj (CudnnConvolution) [TRT] Tactic: 0 Time: 0.316329 [TRT] Tactic: 1 Time: 0.274244 [TRT] Tactic: 2 Time: 0.535912 [TRT] Tactic: 4 skipped. Scratch requested: 247709696, available: 33554432 [TRT] Tactic: 5 Time: 9.55792 [TRT] Fastest Tactic: 1 Time: 0.274244 [TRT] --------------- Timing Runner: inception_5a/pool_proj + inception_5a/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5a/pool_proj + inception_5a/relu_pool_proj (CaskConvolution) [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.264193 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v0 Tactic: 1698681053543049347 [TRT] Tactic: 1698681053543049347 Time: 0.227917 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.205754 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.196771 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.215156 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v1 Tactic: 5326823351883942011 [TRT] Tactic: 5326823351883942011 Time: 0.201277 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v0 Tactic: 5500448035057547314 [TRT] Tactic: 5500448035057547314 Time: 0.244844 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.207813 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.238594 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.217057 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v1 Tactic: -6576203419454146580 [TRT] Tactic: -6576203419454146580 Time: 0.230573 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.251354 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.245391 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.257917 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.248438 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.198203 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v0 Tactic: -377491875521947884 [TRT] Tactic: -377491875521947884 Time: 0.214505 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v1 Tactic: -37215280111360163 [TRT] Tactic: -37215280111360163 Time: 0.191875 [TRT] Fastest Tactic: -37215280111360163 Time: 0.191875 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 7733247 [TRT] *************** Autotuning format combination: Float(40768,1,5824,832) -> Float(40768,1,5824,832) *************** [TRT] --------------- Timing Runner: inception_5a/pool_proj + inception_5a/relu_pool_proj (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5a/pool_proj + inception_5a/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5a/pool_proj + inception_5a/relu_pool_proj (CaskConvolution) [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 3886731678879822788 [TRT] Tactic: 3886731678879822788 Time: 0.173489 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 6629944304117643200 [TRT] Tactic: 6629944304117643200 Time: 0.226927 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.226745 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.172369 [TRT] Fastest Tactic: -7394439838318485025 Time: 0.172369 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(40768,49,7,1) -> Half(40768,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/pool_proj + inception_5a/relu_pool_proj (CudnnConvolution) [TRT] Tactic: 0 Time: 0.287057 [TRT] Tactic: 1 Time: 0.240677 [TRT] Tactic: 2 Time: 0.451901 [TRT] Tactic: 4 skipped. Scratch requested: 247709696, available: 33554432 [TRT] Tactic: 5 Time: 7.31294 [TRT] Fastest Tactic: 1 Time: 0.240677 [TRT] --------------- Timing Runner: inception_5a/pool_proj + inception_5a/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5a/pool_proj + inception_5a/relu_pool_proj (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 1 [TRT] *************** Autotuning format combination: Half(20384,49:2,7,1) -> Half(40768,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/pool_proj + inception_5a/relu_pool_proj (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] *************** Autotuning format combination: Half(20384,49:2,7,1) -> Half(20384,49:2,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/pool_proj + inception_5a/relu_pool_proj (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.14961 [TRT] Tactic: 655359 Time: 0.167865 [TRT] Tactic: 786431 Time: 0.155078 [TRT] Tactic: 851967 Time: 0.102656 [TRT] Tactic: 1179647 Time: 0.083125 [TRT] Tactic: 1310719 Time: 0.286067 [TRT] Tactic: 1376255 Time: 0.158854 [TRT] Tactic: 1441791 Time: 0.115026 [TRT] Tactic: 1507327 Time: 0.100104 [TRT] Tactic: 1638399 Time: 0.143802 [TRT] Tactic: 1835007 Time: 0.143568 [TRT] Tactic: 1900543 Time: 0.177448 [TRT] Tactic: 2097151 Time: 0.11461 [TRT] Tactic: 2162687 Time: 0.171041 [TRT] Tactic: 2293759 Time: 0.166562 [TRT] Tactic: 2359295 Time: 0.10401 [TRT] Tactic: 2686975 Time: 0.242031 [TRT] Tactic: 3080191 Time: 0.0959115 [TRT] Tactic: 3342335 Time: 0.183437 [TRT] Tactic: 3407871 Time: 0.0920315 [TRT] Tactic: 3538943 Time: 0.093516 [TRT] Tactic: 3670015 Time: 0.312838 [TRT] Tactic: 3932159 Time: 0.0895575 [TRT] Tactic: 3997695 Time: 0.165547 [TRT] Tactic: 4063231 Time: 0.100938 [TRT] Tactic: 4194303 Time: 0.0821355 [TRT] Tactic: 4259839 Time: 0.114427 [TRT] Tactic: 4325375 Time: 0.123307 [TRT] Tactic: 4521983 Time: 0.161927 [TRT] Tactic: 4587519 Time: 0.120885 [TRT] Tactic: 4653055 Time: 0.103073 [TRT] Tactic: 4915199 Time: 0.084792 [TRT] Tactic: 4980735 Time: 0.120703 [TRT] Tactic: 5177343 Time: 0.0839585 [TRT] Tactic: 5242879 Time: 0.0815625 [TRT] Tactic: 5373951 Time: 0.0850785 [TRT] Tactic: 5439487 Time: 0.12211 [TRT] Tactic: 5570559 Time: 0.0954945 [TRT] Tactic: 5636095 Time: 0.10125 [TRT] Tactic: 5701631 Time: 0.148464 [TRT] Tactic: 5767167 Time: 0.107084 [TRT] Tactic: 5832703 Time: 0.0873955 [TRT] Tactic: 5898239 Time: 0.0731515 [TRT] Tactic: 6029311 Time: 0.155338 [TRT] Tactic: 6225919 Time: 0.0806515 [TRT] Tactic: 6291455 Time: 0.0832815 [TRT] Tactic: 6422527 Time: 0.088177 [TRT] Tactic: 6750207 Time: 0.105391 [TRT] Tactic: 6815743 Time: 0.080833 [TRT] Tactic: 6946815 Time: 0.127005 [TRT] Tactic: 7012351 Time: 0.114427 [TRT] Tactic: 7077887 Time: 0.086953 [TRT] Tactic: 7143423 Time: 0.113854 [TRT] Tactic: 7208959 Time: 0.08724 [TRT] Tactic: 7340031 Time: 0.075807 [TRT] Tactic: 7405567 Time: 0.097448 [TRT] Tactic: 7536639 Time: 0.143151 [TRT] Tactic: 7602175 Time: 0.126302 [TRT] Tactic: 7733247 Time: 0.0684115 [TRT] Tactic: 7798783 Time: 0.156172 [TRT] Tactic: 8191999 Time: 0.133646 [TRT] Tactic: 8257535 Time: 0.0852345 [TRT] Tactic: 8323071 Time: 0.119453 [TRT] Tactic: 8650751 Time: 0.12711 [TRT] Tactic: 8716287 Time: 0.0821875 [TRT] Tactic: 9109503 Time: 0.117916 [TRT] Tactic: 9568255 Time: 0.086042 [TRT] Tactic: 9895935 Time: 0.080963 [TRT] Tactic: 10223615 Time: 0.242604 [TRT] Tactic: 10354687 Time: 0.114167 [TRT] Tactic: 10551295 Time: 0.100937 [TRT] Tactic: 10747903 Time: 0.0628905 [TRT] Tactic: 10944511 Time: 0.119375 [TRT] Fastest Tactic: 10747903 Time: 0.0628905 [TRT] --------------- Timing Runner: inception_5a/pool_proj + inception_5a/relu_pool_proj (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5a/pool_proj + inception_5a/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5a/pool_proj + inception_5a/relu_pool_proj (CaskConvolution) [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_interior_nn_v1 Tactic: 3066127711859985668 [TRT] Tactic: 3066127711859985668 Time: 0.098698 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.120912 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.10388 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.094687 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] Tactic: 8163473458334948789 Time: 0.091198 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.0872135 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.092266 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.092422 [TRT] inception_5a/pool_proj + inception_5a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_interior_nn_v1 Tactic: -1716393687483585322 [TRT] Tactic: -1716393687483585322 Time: 0.0897915 [TRT] Fastest Tactic: -4212163711445252890 Time: 0.0872135 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 10747903 [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Float(40768,1,5824,832) *************** [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Half(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Half(20384,49:2,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Float(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Half(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Half(20384,49:2,7,1) *************** [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Float(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Float(40768,1,5824,832) *************** [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Half(20384,49:2,7,1) *************** [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Float(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Float(40768,1,5824,832) *************** [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Half(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Float(21952,49,7,1) -> Float(40768,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 1.07826 [TRT] Tactic: 0 Time: 0.006849 [TRT] Fastest Tactic: 0 Time: 0.006849 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(21952,49,7,1) -> Float(40768,1,5824,832) *************** [TRT] --------------- Timing Runner: inception_5a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0240885 [TRT] Tactic: 0 Time: 0.021901 [TRT] Fastest Tactic: 0 Time: 0.021901 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(21952,49,7,1) -> Half(40768,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.47125 [TRT] Tactic: 0 Time: 0.0219015 [TRT] Fastest Tactic: 0 Time: 0.0219015 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(21952,49,7,1) -> Half(20384,49:2,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0277345 [TRT] Tactic: 0 Time: 0.0242965 [TRT] Fastest Tactic: 0 Time: 0.0242965 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(21952,1,3136,448) -> Float(40768,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.028776 [TRT] Tactic: 0 Time: 0.021875 [TRT] Fastest Tactic: 0 Time: 0.021875 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(21952,1,3136,448) -> Float(40768,1,5824,832) *************** [TRT] --------------- Timing Runner: inception_5a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0195055 [TRT] Tactic: 0 Time: 0.0219015 [TRT] Fastest Tactic: 1002 Time: 0.0195055 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(21952,1,3136,448) -> Half(40768,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0208075 [TRT] Tactic: 0 Time: 0.0216145 [TRT] Fastest Tactic: 1002 Time: 0.0208075 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(21952,1,3136,448) -> Half(20384,49:2,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.025521 [TRT] Tactic: 0 Time: 0.0242185 [TRT] Fastest Tactic: 0 Time: 0.0242185 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(21952,49,7,1) -> Float(40768,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.482735 [TRT] Tactic: 0 Time: 0.0218225 [TRT] Fastest Tactic: 0 Time: 0.0218225 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(21952,49,7,1) -> Float(40768,1,5824,832) *************** [TRT] --------------- Timing Runner: inception_5a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.019661 [TRT] Tactic: 0 Time: 0.022135 [TRT] Fastest Tactic: 1002 Time: 0.019661 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(21952,49,7,1) -> Half(40768,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.480573 [TRT] Tactic: 0 Time: 0.005443 [TRT] Fastest Tactic: 0 Time: 0.005443 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(21952,49,7,1) -> Half(20384,49:2,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.02875 [TRT] Tactic: 0 Time: 0.0147135 [TRT] Fastest Tactic: 0 Time: 0.0147135 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(10976,49:2,7,1) -> Float(40768,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.026589 [TRT] Tactic: 0 Time: 0.024088 [TRT] Fastest Tactic: 0 Time: 0.024088 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(10976,49:2,7,1) -> Float(40768,1,5824,832) *************** [TRT] --------------- Timing Runner: inception_5a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.019531 [TRT] Tactic: 0 Time: 0.0263545 [TRT] Fastest Tactic: 1002 Time: 0.019531 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(10976,49:2,7,1) -> Half(40768,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0335155 [TRT] Tactic: 0 Time: 0.021849 [TRT] Fastest Tactic: 0 Time: 0.021849 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(10976,49:2,7,1) -> Half(20384,49:2,7,1) *************** [TRT] --------------- Timing Runner: inception_5a/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.028854 [TRT] Tactic: 0 Time: 0.005599 [TRT] Fastest Tactic: 0 Time: 0.005599 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Float(40768,1,5824,832) *************** [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Half(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Half(20384,49:2,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Float(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Half(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Half(20384,49:2,7,1) *************** [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Float(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Float(40768,1,5824,832) *************** [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Half(20384,49:2,7,1) *************** [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Float(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Float(40768,1,5824,832) *************** [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Half(40768,49,7,1) *************** [TRT] *************** Autotuning format combination: Float(40768,49,7,1) -> Float(30576,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce (FusedConvActConvolution) [TRT] FusedConvActConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce (CudnnConvolution) [TRT] Tactic: 0 Time: 0.920364 [TRT] Tactic: 1 Time: 0.782865 [TRT] Tactic: 2 Time: 0.878906 [TRT] Tactic: 4 skipped. Scratch requested: 1200156672, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 71993344, available: 33554432 [TRT] Fastest Tactic: 1 Time: 0.782865 [TRT] --------------- Timing Runner: inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce (CaskConvolution) [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.929323 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v0 Tactic: 1698681053543049347 [TRT] Tactic: 1698681053543049347 Time: 0.869115 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.74125 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.718672 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.74487 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v1 Tactic: 5326823351883942011 [TRT] Tactic: 5326823351883942011 Time: 0.708516 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v0 Tactic: 5500448035057547314 [TRT] Tactic: 5500448035057547314 Time: 0.805495 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.735443 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.913333 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.761016 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v1 Tactic: -6576203419454146580 [TRT] Tactic: -6576203419454146580 Time: 0.822682 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.955182 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.857265 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.838776 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.820234 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.72099 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v0 Tactic: -377491875521947884 [TRT] Tactic: -377491875521947884 Time: 0.73552 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v1 Tactic: -37215280111360163 [TRT] Tactic: -37215280111360163 Time: 0.704947 [TRT] Fastest Tactic: -37215280111360163 Time: 0.704947 [TRT] Setting workspace to 71993344enables more tactics for profiling [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -37215280111360163 [TRT] *************** Autotuning format combination: Float(40768,1,5824,832) -> Float(30576,1,4368,624) *************** [TRT] --------------- Timing Runner: inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce (CaskConvolution) [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 3886731678879822788 [TRT] Tactic: 3886731678879822788 Time: 0.6975 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 6629944304117643200 [TRT] Tactic: 6629944304117643200 Time: 0.954401 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.955885 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.694349 [TRT] Fastest Tactic: -7394439838318485025 Time: 0.694349 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(40768,49,7,1) -> Half(30576,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce (CudnnConvolution) [TRT] Tactic: 0 Time: 0.945287 [TRT] Tactic: 1 Time: 0.850312 [TRT] Tactic: 2 Time: 0.841849 [TRT] Tactic: 4 skipped. Scratch requested: 1200156672, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 71993344, available: 33554432 [TRT] Fastest Tactic: 2 Time: 0.841849 [TRT] --------------- Timing Runner: inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] Setting workspace to 71993344enables more tactics for profiling [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 2 [TRT] *************** Autotuning format combination: Half(20384,49:2,7,1) -> Half(30576,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] *************** Autotuning format combination: Half(20384,49:2,7,1) -> Half(15288,49:2,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce (FusedConvActConvolution) [TRT] FusedConvActConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce (CaskConvolution) [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_interior_nn_v1 Tactic: 3066127711859985668 [TRT] Tactic: 3066127711859985668 Time: 0.418593 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.475442 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.442734 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.380964 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] Tactic: 8163473458334948789 Time: 0.358568 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.367552 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.374245 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.366406 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_interior_nn_v1 Tactic: -1716393687483585322 [TRT] Tactic: -1716393687483585322 Time: 0.359271 [TRT] Fastest Tactic: 8163473458334948789 Time: 0.358568 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: 8163473458334948789 [TRT] *************** Autotuning Reformat:Float(30576,49,7,1) -> Float(30576,1,4368,624) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0460935 [TRT] Tactic: 0 Time: 0.049896 [TRT] Fastest Tactic: 1002 Time: 0.0460935 [TRT] *************** Autotuning Reformat:Float(30576,49,7,1) -> Half(30576,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.943828 [TRT] Tactic: 0 Time: 0.0349485 [TRT] Fastest Tactic: 0 Time: 0.0349485 [TRT] *************** Autotuning Reformat:Float(30576,49,7,1) -> Half(15288,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0577865 [TRT] Tactic: 0 Time: 0.028698 [TRT] Fastest Tactic: 0 Time: 0.028698 [TRT] *************** Autotuning Reformat:Float(30576,1,4368,624) -> Float(30576,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0592445 [TRT] Tactic: 0 Time: 0.046146 [TRT] Fastest Tactic: 0 Time: 0.046146 [TRT] *************** Autotuning Reformat:Float(30576,1,4368,624) -> Half(30576,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.045 [TRT] Tactic: 0 Time: 0.0473175 [TRT] Fastest Tactic: 1002 Time: 0.045 [TRT] *************** Autotuning Reformat:Float(30576,1,4368,624) -> Half(15288,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.05099 [TRT] Tactic: 0 Time: 0.0542705 [TRT] Fastest Tactic: 1002 Time: 0.05099 [TRT] *************** Autotuning Reformat:Half(30576,49,7,1) -> Float(30576,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.973802 [TRT] Tactic: 0 Time: 0.03099 [TRT] Fastest Tactic: 0 Time: 0.03099 [TRT] *************** Autotuning Reformat:Half(30576,49,7,1) -> Float(30576,1,4368,624) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0417965 [TRT] Tactic: 0 Time: 0.0503125 [TRT] Fastest Tactic: 1002 Time: 0.0417965 [TRT] *************** Autotuning Reformat:Half(30576,49,7,1) -> Half(15288,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0625 [TRT] Tactic: 0 Time: 0.028698 [TRT] Fastest Tactic: 0 Time: 0.028698 [TRT] *************** Autotuning Reformat:Half(15288,49:2,7,1) -> Float(30576,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0545315 [TRT] Tactic: 0 Time: 0.0254945 [TRT] Fastest Tactic: 0 Time: 0.0254945 [TRT] *************** Autotuning Reformat:Half(15288,49:2,7,1) -> Float(30576,1,4368,624) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0432035 [TRT] Tactic: 0 Time: 0.0579685 [TRT] Fastest Tactic: 1002 Time: 0.0432035 [TRT] *************** Autotuning Reformat:Half(15288,49:2,7,1) -> Half(30576,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.068125 [TRT] Tactic: 0 Time: 0.0240625 [TRT] Fastest Tactic: 0 Time: 0.0240625 [TRT] *************** Autotuning Reformat:Float(30576,49,7,1) -> Float(30576,1,4368,624) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.014948 [TRT] Tactic: 0 Time: 0.0173435 [TRT] Fastest Tactic: 1002 Time: 0.014948 [TRT] *************** Autotuning Reformat:Float(30576,49,7,1) -> Half(30576,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.355443 [TRT] Tactic: 0 Time: 0.0172655 [TRT] Fastest Tactic: 0 Time: 0.0172655 [TRT] *************** Autotuning Reformat:Float(30576,49,7,1) -> Half(15288,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.021849 [TRT] Tactic: 0 Time: 0.0194015 [TRT] Fastest Tactic: 0 Time: 0.0194015 [TRT] *************** Autotuning Reformat:Float(30576,1,4368,624) -> Float(30576,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0195575 [TRT] Tactic: 0 Time: 0.017084 [TRT] Fastest Tactic: 0 Time: 0.017084 [TRT] *************** Autotuning Reformat:Float(30576,1,4368,624) -> Half(30576,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0172135 [TRT] Tactic: 0 Time: 0.017136 [TRT] Fastest Tactic: 0 Time: 0.017136 [TRT] *************** Autotuning Reformat:Float(30576,1,4368,624) -> Half(15288,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0209635 [TRT] Tactic: 0 Time: 0.0194535 [TRT] Fastest Tactic: 0 Time: 0.0194535 [TRT] *************** Autotuning Reformat:Half(30576,49,7,1) -> Float(30576,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.363697 [TRT] Tactic: 0 Time: 0.012448 [TRT] Fastest Tactic: 0 Time: 0.012448 [TRT] *************** Autotuning Reformat:Half(30576,49,7,1) -> Float(30576,1,4368,624) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0162495 [TRT] Tactic: 0 Time: 0.0184375 [TRT] Fastest Tactic: 1002 Time: 0.0162495 [TRT] *************** Autotuning Reformat:Half(30576,49,7,1) -> Half(15288,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0233855 [TRT] Tactic: 0 Time: 0.010417 [TRT] Fastest Tactic: 0 Time: 0.010417 [TRT] *************** Autotuning Reformat:Half(15288,49:2,7,1) -> Float(30576,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0221095 [TRT] Tactic: 0 Time: 0.0101825 [TRT] Fastest Tactic: 0 Time: 0.0101825 [TRT] *************** Autotuning Reformat:Half(15288,49:2,7,1) -> Float(30576,1,4368,624) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0172135 [TRT] Tactic: 0 Time: 0.0197135 [TRT] Fastest Tactic: 1002 Time: 0.0172135 [TRT] *************** Autotuning Reformat:Half(15288,49:2,7,1) -> Half(30576,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.02875 [TRT] Tactic: 0 Time: 0.017135 [TRT] Fastest Tactic: 0 Time: 0.017135 [TRT] *************** Autotuning format combination: Float(30576,49,7,1) -> Float(50176,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/3x3 + inception_5b/relu_3x3 (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5b/3x3 + inception_5b/relu_3x3 (FusedConvActConvolution) [TRT] Tactic: 524287 Time: 0.491146 [TRT] Tactic: 720895 Time: 0.890989 [TRT] Tactic: 983039 Time: 0.428802 [TRT] Tactic: 1048575 Time: 0.643229 [TRT] Tactic: 1703935 Time: 0.620755 [TRT] Tactic: 1769471 Time: 0.563672 [TRT] Tactic: 1966079 Time: 0.470547 [TRT] Tactic: 2031615 Time: 0.787501 [TRT] Tactic: 2228223 Time: 0.739636 [TRT] Tactic: 2424831 Time: 0.823099 [TRT] Tactic: 2621439 Time: 0.707474 [TRT] Tactic: 2752511 Time: 0.56513 [TRT] Tactic: 2818047 Time: 0.728099 [TRT] Tactic: 2883583 Time: 0.603724 [TRT] Tactic: 3014655 Time: 0.430625 [TRT] Tactic: 3145727 Time: 0.535234 [TRT] Tactic: 3473407 Time: 0.599635 [TRT] Tactic: 3604479 Time: 0.420651 [TRT] Tactic: 3735551 Time: 0.779844 [TRT] Tactic: 4390911 Time: 0.530261 [TRT] Tactic: 5046271 Time: 0.470182 [TRT] Tactic: 5963775 Time: 0.430755 [TRT] Tactic: 6160383 Time: 0.465729 [TRT] Tactic: 6488063 Time: 0.418802 [TRT] Tactic: 6881279 Time: 0.528047 [TRT] Tactic: 7274495 Time: 0.636355 [TRT] Tactic: 7864319 Time: 0.738594 [TRT] Tactic: 7995391 Time: 0.517761 [TRT] Tactic: 8585215 Time: 0.552552 [TRT] Tactic: 8847359 Time: 0.419974 [TRT] Tactic: 8978431 Time: 0.435443 [TRT] Tactic: 9043967 Time: 0.397135 [TRT] Tactic: 9175039 Time: 0.420625 [TRT] Tactic: 9502719 Time: 0.559453 [TRT] Tactic: 9830399 Time: 0.489011 [TRT] Tactic: 9961471 Time: 0.460938 [TRT] Tactic: 10027007 Time: 0.547761 [TRT] Tactic: 10092543 Time: 0.529661 [TRT] Tactic: 10289151 Time: 0.469609 [TRT] Tactic: 10485759 Time: 0.539791 [TRT] Tactic: 10682367 Time: 0.71612 [TRT] Tactic: 10813439 Time: 0.489479 [TRT] Fastest Tactic: 9043967 Time: 0.397135 [TRT] --------------- Timing Runner: inception_5b/3x3 + inception_5b/relu_3x3 (CudnnConvolution) [TRT] Tactic: 0 Time: 0.867526 [TRT] Tactic: 1 Time: 1.04565 [TRT] Tactic: 2 Time: 0.908542 [TRT] Tactic: 4 skipped. Scratch requested: 172965888, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 323371008, available: 33554432 [TRT] Tactic: 6 Time: 1.00359 [TRT] Fastest Tactic: 0 Time: 0.867526 [TRT] --------------- Timing Runner: inception_5b/3x3 + inception_5b/relu_3x3 (CaskConvolution) [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 1.1431 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v0 Tactic: 1754984623894446479 [TRT] Tactic: 1754984623894446479 Time: 1.47039 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v0 Tactic: 3611739942397549984 [TRT] Tactic: 3611739942397549984 Time: 0.95513 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148n_nt_v1 Tactic: 3827454225649558724 [TRT] Tactic: 3827454225649558724 Time: 0.674401 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v1 Tactic: 4337000649858996379 [TRT] Tactic: 4337000649858996379 Time: 0.994609 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.957526 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.894661 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.929479 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 5921334924264294896 [TRT] Tactic: 5921334924264294896 Time: 0.530365 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.965234 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 1.22148 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148t_nt_v1 Tactic: 7852627285308570038 [TRT] Tactic: 7852627285308570038 Time: 0.639766 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v1 Tactic: -9137461792520977713 [TRT] Tactic: -9137461792520977713 Time: 0.973359 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148n_nt_v0 Tactic: -8776506421218919509 [TRT] Tactic: -8776506421218919509 Time: 0.629791 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.987969 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v0 Tactic: -8133971918129952780 [TRT] Tactic: -8133971918129952780 Time: 1.12505 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v1 Tactic: -6092040395344634144 [TRT] Tactic: -6092040395344634144 Time: 1.19094 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 1.45362 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.978516 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_relu_tile148t_nt_v0 Tactic: -2318106587342035239 [TRT] Tactic: -2318106587342035239 Time: 0.629375 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_winograd_128x128_ldg1_ldg4_mobile_relu_tile148t_nt_v0 Tactic: -1343271414618805657 [TRT] Tactic: -1343271414618805657 Time: 0.477369 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 1.11544 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 1.06125 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.896953 [TRT] Fastest Tactic: -1343271414618805657 Time: 0.477369 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 9043967 [TRT] *************** Autotuning format combination: Float(30576,1,4368,624) -> Float(50176,1,7168,1024) *************** [TRT] --------------- Timing Runner: inception_5b/3x3 + inception_5b/relu_3x3 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5b/3x3 + inception_5b/relu_3x3 (CaskConvolution) [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 1.01802 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.833619 [TRT] Fastest Tactic: -7394439838318485025 Time: 0.833619 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(30576,49,7,1) -> Half(50176,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/3x3 + inception_5b/relu_3x3 (CudnnConvolution) [TRT] Tactic: 0 Time: 0.853594 [TRT] Tactic: 1 Time: 1.1905 [TRT] Tactic: 2 Time: 0.886302 [TRT] Tactic: 4 skipped. Scratch requested: 172965888, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 323371008, available: 33554432 [TRT] Tactic: 6 Time: 1.94422 [TRT] Fastest Tactic: 0 Time: 0.853594 [TRT] --------------- Timing Runner: inception_5b/3x3 + inception_5b/relu_3x3 (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 0 [TRT] *************** Autotuning format combination: Half(15288,49:2,7,1) -> Half(25088,49:2,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/3x3 + inception_5b/relu_3x3 (FusedConvActConvolution) [TRT] Tactic: 524287 Time: 0.262109 [TRT] Tactic: 720895 Time: 0.513047 [TRT] Tactic: 983039 Time: 0.241224 [TRT] Tactic: 1048575 Time: 0.354896 [TRT] Tactic: 1703935 Time: 0.348932 [TRT] Tactic: 1769471 Time: 1.9824 [TRT] Tactic: 1966079 Time: 0.279869 [TRT] Tactic: 2031615 Time: 0.431953 [TRT] Tactic: 2228223 Time: 0.430781 [TRT] Tactic: 2424831 Time: 0.619792 [TRT] Tactic: 2621439 Time: 0.38263 [TRT] Tactic: 2752511 Time: 0.316875 [TRT] Tactic: 2818047 Time: 0.400833 [TRT] Tactic: 2883583 Time: 0.37263 [TRT] Tactic: 3014655 Time: 0.250781 [TRT] Tactic: 3145727 Time: 0.299323 [TRT] Tactic: 3473407 Time: 0.368751 [TRT] Tactic: 3604479 Time: 0.243881 [TRT] Tactic: 3735551 Time: 0.421719 [TRT] Tactic: 4390911 Time: 0.313281 [TRT] Tactic: 5046271 Time: 0.256225 [TRT] Tactic: 5963775 Time: 0.254531 [TRT] Tactic: 6160383 Time: 0.286432 [TRT] Tactic: 6488063 Time: 0.237005 [TRT] Tactic: 6881279 Time: 0.297135 [TRT] Tactic: 7274495 Time: 0.366302 [TRT] Tactic: 7864319 Time: 0.411927 [TRT] Tactic: 7995391 Time: 0.298541 [TRT] Tactic: 8585215 Time: 0.320417 [TRT] Tactic: 8847359 Time: 0.231016 [TRT] Tactic: 8978431 Time: 0.252265 [TRT] Tactic: 9043967 Time: 0.218386 [TRT] Tactic: 9175039 Time: 0.246719 [TRT] Tactic: 9502719 Time: 0.301432 [TRT] Tactic: 9830399 Time: 0.268463 [TRT] Tactic: 9961471 Time: 0.263125 [TRT] Tactic: 10027007 Time: 0.300651 [TRT] Tactic: 10092543 Time: 0.311927 [TRT] Tactic: 10289151 Time: 0.277942 [TRT] Tactic: 10485759 Time: 0.295781 [TRT] Tactic: 10682367 Time: 0.377917 [TRT] Tactic: 10813439 Time: 0.284401 [TRT] Fastest Tactic: 9043967 Time: 0.218386 [TRT] --------------- Timing Runner: inception_5b/3x3 + inception_5b/relu_3x3 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5b/3x3 + inception_5b/relu_3x3 (CaskConvolution) [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.601042 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_large_nn_v1 Tactic: 3650389455493082349 [TRT] Tactic: 3650389455493082349 Time: 0.628255 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_winograd_fp16x2_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 4772821744921268633 [TRT] Tactic: 4772821744921268633 Time: 0.27961 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.491145 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.491354 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_large_nn_v1 Tactic: -6490690591794140522 [TRT] Tactic: -6490690591794140522 Time: 0.50164 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_large_nn_v1 Tactic: -4686027666808657977 [TRT] Tactic: -4686027666808657977 Time: 0.486693 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.445755 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.477344 [TRT] inception_5b/3x3 + inception_5b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.453828 [TRT] Fastest Tactic: 4772821744921268633 Time: 0.27961 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 9043967 [TRT] *************** Autotuning Reformat:Float(50176,49,7,1) -> Float(50176,1,7168,1024) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.026745 [TRT] Tactic: 0 Time: 0.031536 [TRT] Fastest Tactic: 1002 Time: 0.026745 [TRT] *************** Autotuning Reformat:Float(50176,49,7,1) -> Half(50176,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.703073 [TRT] Tactic: 0 Time: 0.0291145 [TRT] Fastest Tactic: 0 Time: 0.0291145 [TRT] *************** Autotuning Reformat:Float(50176,49,7,1) -> Half(25088,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0383335 [TRT] Tactic: 0 Time: 0.019375 [TRT] Fastest Tactic: 0 Time: 0.019375 [TRT] *************** Autotuning Reformat:Float(50176,1,7168,1024) -> Float(50176,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0357815 [TRT] Tactic: 0 Time: 0.028776 [TRT] Fastest Tactic: 0 Time: 0.028776 [TRT] *************** Autotuning Reformat:Float(50176,1,7168,1024) -> Half(50176,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.02875 [TRT] Tactic: 0 Time: 0.031146 [TRT] Fastest Tactic: 1002 Time: 0.02875 [TRT] *************** Autotuning Reformat:Float(50176,1,7168,1024) -> Half(25088,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.033385 [TRT] Tactic: 0 Time: 0.035625 [TRT] Fastest Tactic: 1002 Time: 0.033385 [TRT] *************** Autotuning Reformat:Half(50176,49,7,1) -> Float(50176,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.720338 [TRT] Tactic: 0 Time: 0.0302345 [TRT] Fastest Tactic: 0 Time: 0.0302345 [TRT] *************** Autotuning Reformat:Half(50176,49,7,1) -> Float(50176,1,7168,1024) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0266145 [TRT] Tactic: 0 Time: 0.03237 [TRT] Fastest Tactic: 1002 Time: 0.0266145 [TRT] *************** Autotuning Reformat:Half(50176,49,7,1) -> Half(25088,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.040469 [TRT] Tactic: 0 Time: 0.019479 [TRT] Fastest Tactic: 0 Time: 0.019479 [TRT] *************** Autotuning Reformat:Half(25088,49:2,7,1) -> Float(50176,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.035885 [TRT] Tactic: 0 Time: 0.0353125 [TRT] Fastest Tactic: 0 Time: 0.0353125 [TRT] *************** Autotuning Reformat:Half(25088,49:2,7,1) -> Float(50176,1,7168,1024) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.026953 [TRT] Tactic: 0 Time: 0.0379945 [TRT] Fastest Tactic: 1002 Time: 0.026953 [TRT] *************** Autotuning Reformat:Half(25088,49:2,7,1) -> Half(50176,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.045182 [TRT] Tactic: 0 Time: 0.031068 [TRT] Fastest Tactic: 0 Time: 0.031068 [TRT] *************** Autotuning Reformat:Float(30576,49,7,1) -> Float(30576,1,4368,624) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.014974 [TRT] Tactic: 0 Time: 0.007864 [TRT] Fastest Tactic: 0 Time: 0.007864 [TRT] *************** Autotuning Reformat:Float(30576,49,7,1) -> Half(30576,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.093698 [TRT] Tactic: 0 Time: 0.007812 [TRT] Fastest Tactic: 0 Time: 0.007812 [TRT] *************** Autotuning Reformat:Float(30576,49,7,1) -> Half(15288,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.01013 [TRT] Tactic: 0 Time: 0.007891 [TRT] Fastest Tactic: 0 Time: 0.007891 [TRT] *************** Autotuning Reformat:Float(30576,1,4368,624) -> Float(30576,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.014974 [TRT] Tactic: 0 Time: 0.007943 [TRT] Fastest Tactic: 0 Time: 0.007943 [TRT] *************** Autotuning Reformat:Float(30576,1,4368,624) -> Half(30576,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0102605 [TRT] Tactic: 0 Time: 0.0078645 [TRT] Fastest Tactic: 0 Time: 0.0078645 [TRT] *************** Autotuning Reformat:Float(30576,1,4368,624) -> Half(15288,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0103125 [TRT] Tactic: 0 Time: 0.0079425 [TRT] Fastest Tactic: 0 Time: 0.0079425 [TRT] *************** Autotuning Reformat:Half(30576,49,7,1) -> Float(30576,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.095911 [TRT] Tactic: 0 Time: 0.005573 [TRT] Fastest Tactic: 0 Time: 0.005573 [TRT] *************** Autotuning Reformat:Half(30576,49,7,1) -> Float(30576,1,4368,624) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.009635 [TRT] Tactic: 0 Time: 0.007917 [TRT] Fastest Tactic: 0 Time: 0.007917 [TRT] *************** Autotuning Reformat:Half(30576,49,7,1) -> Half(15288,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.012656 [TRT] Tactic: 0 Time: 0.0055205 [TRT] Fastest Tactic: 0 Time: 0.0055205 [TRT] *************** Autotuning Reformat:Half(15288,49:2,7,1) -> Float(30576,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0103125 [TRT] Tactic: 0 Time: 0.0055985 [TRT] Fastest Tactic: 0 Time: 0.0055985 [TRT] *************** Autotuning Reformat:Half(15288,49:2,7,1) -> Float(30576,1,4368,624) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0103125 [TRT] Tactic: 0 Time: 0.0078385 [TRT] Fastest Tactic: 0 Time: 0.0078385 [TRT] *************** Autotuning Reformat:Half(15288,49:2,7,1) -> Half(30576,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.012604 [TRT] Tactic: 0 Time: 0.007942 [TRT] Fastest Tactic: 0 Time: 0.007942 [TRT] *************** Autotuning format combination: Float(30576,49,7,1) -> Float(50176,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/5x5 + inception_5b/relu_5x5 (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5b/5x5 + inception_5b/relu_5x5 (FusedConvActConvolution) [TRT] Tactic: 393215 Time: 0.146484 [TRT] Tactic: 917503 Time: 0.202369 [TRT] Tactic: 1114111 Time: 0.172448 [TRT] Tactic: 1245183 Time: 0.203516 [TRT] Tactic: 1572863 Time: 0.219427 [TRT] Tactic: 2490367 Time: 0.189714 [TRT] Tactic: 2555903 Time: 0.229661 [TRT] Tactic: 2949119 Time: 0.200287 [TRT] Tactic: 3211263 Time: 0.183151 [TRT] Tactic: 3801087 Time: 0.209765 [TRT] Tactic: 3866623 Time: 0.169531 [TRT] Tactic: 4128767 Time: 0.186927 [TRT] Tactic: 4456447 Time: 0.192708 [TRT] Tactic: 4718591 Time: 0.19789 [TRT] Tactic: 4784127 Time: 0.279896 [TRT] Tactic: 4849663 Time: 0.169089 [TRT] Tactic: 5111807 Time: 0.172213 [TRT] Tactic: 5308415 Time: 0.204297 [TRT] Tactic: 5505023 Time: 0.198619 [TRT] Tactic: 6094847 Time: 0.199454 [TRT] Tactic: 6356991 Time: 0.229557 [TRT] Tactic: 6553599 Time: 0.147344 [TRT] Tactic: 6619135 Time: 0.162604 [TRT] Tactic: 6684671 Time: 0.164348 [TRT] Tactic: 7471103 Time: 0.172213 [TRT] Tactic: 7667711 Time: 0.194974 [TRT] Tactic: 7929855 Time: 0.167448 [TRT] Tactic: 8060927 Time: 0.16349 [TRT] Tactic: 8126463 Time: 0.21599 [TRT] Tactic: 8388607 Time: 0.149661 [TRT] Tactic: 8519679 Time: 0.235469 [TRT] Tactic: 8781823 Time: 0.179583 [TRT] Tactic: 8912895 Time: 0.212578 [TRT] Tactic: 9240575 Time: 0.194869 [TRT] Tactic: 9306111 Time: 0.207969 [TRT] Tactic: 9371647 Time: 0.147344 [TRT] Tactic: 9437183 Time: 0.199843 [TRT] Tactic: 9633791 Time: 0.168984 [TRT] Tactic: 9699327 Time: 0.150208 [TRT] Tactic: 9764863 Time: 0.188437 [TRT] Tactic: 10158079 Time: 0.142916 [TRT] Tactic: 10420223 Time: 0.17276 [TRT] Tactic: 10616831 Time: 0.209219 [TRT] Tactic: 10878975 Time: 0.195443 [TRT] Fastest Tactic: 10158079 Time: 0.142916 [TRT] --------------- Timing Runner: inception_5b/5x5 + inception_5b/relu_5x5 (CudnnConvolution) [TRT] Tactic: 0 Time: 0.345313 [TRT] Tactic: 1 Time: 0.336224 [TRT] Tactic: 2 Time: 0.419973 [TRT] Tactic: 4 Time: 2.12815 [TRT] Tactic: 5 Time: 9.81 [TRT] Fastest Tactic: 1 Time: 0.336224 [TRT] --------------- Timing Runner: inception_5b/5x5 + inception_5b/relu_5x5 (CaskConvolution) [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.378541 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v0 Tactic: 1754984623894446479 [TRT] Tactic: 1754984623894446479 Time: 0.409584 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v0 Tactic: 3611739942397549984 [TRT] Tactic: 3611739942397549984 Time: 0.290469 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v1 Tactic: 4337000649858996379 [TRT] Tactic: 4337000649858996379 Time: 0.305313 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.275807 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.241666 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.263568 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.296198 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.287396 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_large_nn_v1 Tactic: -9137461792520977713 [TRT] Tactic: -9137461792520977713 Time: 0.303489 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.302162 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_large_nn_v0 Tactic: -8133971918129952780 [TRT] Tactic: -8133971918129952780 Time: 0.363932 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_large_nn_v1 Tactic: -6092040395344634144 [TRT] Tactic: -6092040395344634144 Time: 0.403125 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.404765 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.275339 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.363959 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.304193 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.243437 [TRT] Fastest Tactic: 5137655947464784826 Time: 0.241666 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 10158079 [TRT] *************** Autotuning format combination: Float(30576,1,4368,624) -> Float(50176,1,7168,1024) *************** [TRT] --------------- Timing Runner: inception_5b/5x5 + inception_5b/relu_5x5 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5b/5x5 + inception_5b/relu_5x5 (CaskConvolution) [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.327109 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.204609 [TRT] Fastest Tactic: -7394439838318485025 Time: 0.204609 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(30576,49,7,1) -> Half(50176,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/5x5 + inception_5b/relu_5x5 (CudnnConvolution) [TRT] Tactic: 0 Time: 0.337448 [TRT] Tactic: 1 Time: 0.672969 [TRT] Tactic: 2 Time: 0.432448 [TRT] Tactic: 4 Time: 2.31081 [TRT] Tactic: 5 Time: 9.53727 [TRT] Fastest Tactic: 0 Time: 0.337448 [TRT] --------------- Timing Runner: inception_5b/5x5 + inception_5b/relu_5x5 (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 0 [TRT] *************** Autotuning format combination: Half(15288,49:2,7,1) -> Half(25088,49:2,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/5x5 + inception_5b/relu_5x5 (FusedConvActConvolution) [TRT] Tactic: 393215 Time: 0.0814585 [TRT] Tactic: 917503 Time: 0.117604 [TRT] Tactic: 1114111 Time: 0.0877085 [TRT] Tactic: 1245183 Time: 0.108386 [TRT] Tactic: 1572863 Time: 0.120703 [TRT] Tactic: 2490367 Time: 0.124557 [TRT] Tactic: 2555903 Time: 0.153125 [TRT] Tactic: 2949119 Time: 0.125443 [TRT] Tactic: 3211263 Time: 0.103516 [TRT] Tactic: 3801087 Time: 0.109374 [TRT] Tactic: 3866623 Time: 0.0929165 [TRT] Tactic: 4128767 Time: 0.102109 [TRT] Tactic: 4456447 Time: 0.112422 [TRT] Tactic: 4718591 Time: 0.088281 [TRT] Tactic: 4784127 Time: 0.156173 [TRT] Tactic: 4849663 Time: 0.094453 [TRT] Tactic: 5111807 Time: 0.0916665 [TRT] Tactic: 5308415 Time: 0.113151 [TRT] Tactic: 5505023 Time: 0.101432 [TRT] Tactic: 6094847 Time: 0.106927 [TRT] Tactic: 6356991 Time: 0.15349 [TRT] Tactic: 6553599 Time: 0.0791665 [TRT] Tactic: 6619135 Time: 0.0860155 [TRT] Tactic: 6684671 Time: 0.0868225 [TRT] Tactic: 7471103 Time: 0.0969015 [TRT] Tactic: 7667711 Time: 0.089453 [TRT] Tactic: 7929855 Time: 0.0780465 [TRT] Tactic: 8060927 Time: 0.086693 [TRT] Tactic: 8126463 Time: 0.101562 [TRT] Tactic: 8388607 Time: 0.0840105 [TRT] Tactic: 8519679 Time: 0.116849 [TRT] Tactic: 8781823 Time: 0.0940105 [TRT] Tactic: 8912895 Time: 0.117005 [TRT] Tactic: 9240575 Time: 0.0902345 [TRT] Tactic: 9306111 Time: 0.107239 [TRT] Tactic: 9371647 Time: 0.076433 [TRT] Tactic: 9437183 Time: 0.125 [TRT] Tactic: 9633791 Time: 0.0922395 [TRT] Tactic: 9699327 Time: 0.078359 [TRT] Tactic: 9764863 Time: 0.105079 [TRT] Tactic: 10158079 Time: 0.080182 [TRT] Tactic: 10420223 Time: 0.111641 [TRT] Tactic: 10616831 Time: 0.109062 [TRT] Tactic: 10878975 Time: 0.111354 [TRT] Fastest Tactic: 9371647 Time: 0.076433 [TRT] --------------- Timing Runner: inception_5b/5x5 + inception_5b/relu_5x5 (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5b/5x5 + inception_5b/relu_5x5 (CaskConvolution) [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.205261 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_large_nn_v1 Tactic: 3650389455493082349 [TRT] Tactic: 3650389455493082349 Time: 0.215182 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.140078 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.152449 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_large_nn_v1 Tactic: -6490690591794140522 [TRT] Tactic: -6490690591794140522 Time: 0.158567 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_large_nn_v1 Tactic: -4686027666808657977 [TRT] Tactic: -4686027666808657977 Time: 0.145443 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.118671 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.134089 [TRT] inception_5b/5x5 + inception_5b/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.130495 [TRT] Fastest Tactic: -4212163711445252890 Time: 0.118671 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 9371647 [TRT] *************** Autotuning Reformat:Float(50176,49,7,1) -> Float(50176,1,7168,1024) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0139065 [TRT] Tactic: 0 Time: 0.012578 [TRT] Fastest Tactic: 0 Time: 0.012578 [TRT] *************** Autotuning Reformat:Float(50176,49,7,1) -> Half(50176,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.239558 [TRT] Tactic: 0 Time: 0.012709 [TRT] Fastest Tactic: 0 Time: 0.012709 [TRT] *************** Autotuning Reformat:Float(50176,49,7,1) -> Half(25088,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.016875 [TRT] Tactic: 0 Time: 0.008021 [TRT] Fastest Tactic: 0 Time: 0.008021 [TRT] *************** Autotuning Reformat:Float(50176,1,7168,1024) -> Float(50176,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0165885 [TRT] Tactic: 0 Time: 0.0128385 [TRT] Fastest Tactic: 0 Time: 0.0128385 [TRT] *************** Autotuning Reformat:Float(50176,1,7168,1024) -> Half(50176,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0126565 [TRT] Tactic: 0 Time: 0.0125525 [TRT] Fastest Tactic: 0 Time: 0.0125525 [TRT] *************** Autotuning Reformat:Float(50176,1,7168,1024) -> Half(25088,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0163025 [TRT] Tactic: 0 Time: 0.0147655 [TRT] Fastest Tactic: 0 Time: 0.0147655 [TRT] *************** Autotuning Reformat:Half(50176,49,7,1) -> Float(50176,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.244453 [TRT] Tactic: 0 Time: 0.012604 [TRT] Fastest Tactic: 0 Time: 0.012604 [TRT] *************** Autotuning Reformat:Half(50176,49,7,1) -> Float(50176,1,7168,1024) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.012604 [TRT] Tactic: 0 Time: 0.012578 [TRT] Fastest Tactic: 0 Time: 0.012578 [TRT] *************** Autotuning Reformat:Half(50176,49,7,1) -> Half(25088,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0195575 [TRT] Tactic: 0 Time: 0.00789 [TRT] Fastest Tactic: 0 Time: 0.00789 [TRT] *************** Autotuning Reformat:Half(25088,49:2,7,1) -> Float(50176,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.017214 [TRT] Tactic: 0 Time: 0.0147915 [TRT] Fastest Tactic: 0 Time: 0.0147915 [TRT] *************** Autotuning Reformat:Half(25088,49:2,7,1) -> Float(50176,1,7168,1024) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.012605 [TRT] Tactic: 0 Time: 0.0148175 [TRT] Fastest Tactic: 1002 Time: 0.012605 [TRT] *************** Autotuning Reformat:Half(25088,49:2,7,1) -> Half(50176,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0195055 [TRT] Tactic: 0 Time: 0.012552 [TRT] Fastest Tactic: 0 Time: 0.012552 [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Half(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Half(20384,49:2,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Float(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Half(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Half(20384,49:2,7,1) *************** [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Float(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Half(20384,49:2,7,1) *************** [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Float(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Half(40768,49,7,1) *************** [TRT] *************** Autotuning format combination: Float(40768,49,7,1) -> Float(40768,49,7,1) *************** [TRT] *************** Autotuning format combination: Half(40768,49,7,1) -> Half(40768,49,7,1) *************** [TRT] *************** Autotuning format combination: Half(20384,49:2,7,1) -> Half(20384,49:2,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Float(40768,1,5824,832) *************** [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Half(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,49,7,1) -> Half(20384,49:2,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Float(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Half(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Float(40768,1,5824,832) -> Half(20384,49:2,7,1) *************** [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Float(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Float(40768,1,5824,832) *************** [TRT] *************** Autotuning Reformat:Half(40768,49,7,1) -> Half(20384,49:2,7,1) *************** [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Float(40768,49,7,1) *************** [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Float(40768,1,5824,832) *************** [TRT] *************** Autotuning Reformat:Half(20384,49:2,7,1) -> Half(40768,49,7,1) *************** [TRT] *************** Autotuning format combination: Float(40768,49,7,1) -> Float(50176,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/pool_proj + inception_5b/relu_pool_proj (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5b/pool_proj + inception_5b/relu_pool_proj (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.285886 [TRT] Tactic: 655359 Time: 0.1575 [TRT] Tactic: 786431 Time: 0.273099 [TRT] Tactic: 851967 Time: 0.184088 [TRT] Tactic: 1179647 Time: 0.176459 [TRT] Tactic: 1310719 Time: 0.423464 [TRT] Tactic: 1376255 Time: 0.311536 [TRT] Tactic: 1441791 Time: 0.229297 [TRT] Tactic: 1507327 Time: 0.188906 [TRT] Tactic: 1638399 Time: 0.268281 [TRT] Tactic: 1835007 Time: 0.267109 [TRT] Tactic: 1900543 Time: 0.34198 [TRT] Tactic: 2097151 Time: 0.170547 [TRT] Tactic: 2162687 Time: 0.33724 [TRT] Tactic: 2293759 Time: 0.323412 [TRT] Tactic: 2359295 Time: 0.199766 [TRT] Tactic: 2686975 Time: 0.307474 [TRT] Tactic: 3080191 Time: 0.124115 [TRT] Tactic: 3342335 Time: 0.348724 [TRT] Tactic: 3407871 Time: 0.182578 [TRT] Tactic: 3538943 Time: 0.189714 [TRT] Tactic: 3670015 Time: 0.288125 [TRT] Tactic: 3932159 Time: 0.28862 [TRT] Tactic: 3997695 Time: 0.289584 [TRT] Tactic: 4063231 Time: 0.18599 [TRT] Tactic: 4194303 Time: 0.150781 [TRT] Tactic: 4259839 Time: 0.179114 [TRT] Tactic: 4325375 Time: 0.251172 [TRT] Tactic: 4521983 Time: 0.316458 [TRT] Tactic: 4587519 Time: 0.225964 [TRT] Tactic: 4653055 Time: 0.191537 [TRT] Tactic: 4915199 Time: 0.157526 [TRT] Tactic: 4980735 Time: 0.230235 [TRT] Tactic: 5177343 Time: 0.174922 [TRT] Tactic: 5242879 Time: 0.174688 [TRT] Tactic: 5373951 Time: 0.179245 [TRT] Tactic: 5439487 Time: 0.154479 [TRT] Tactic: 5570559 Time: 0.122968 [TRT] Tactic: 5636095 Time: 0.18677 [TRT] Tactic: 5701631 Time: 0.314427 [TRT] Tactic: 5767167 Time: 0.209505 [TRT] Tactic: 5832703 Time: 0.180625 [TRT] Tactic: 5898239 Time: 0.139479 [TRT] Tactic: 6029311 Time: 0.304505 [TRT] Tactic: 6225919 Time: 0.123854 [TRT] Tactic: 6291455 Time: 0.176771 [TRT] Tactic: 6422527 Time: 0.162057 [TRT] Tactic: 6750207 Time: 0.160156 [TRT] Tactic: 6815743 Time: 0.17651 [TRT] Tactic: 6946815 Time: 0.272682 [TRT] Tactic: 7012351 Time: 0.171745 [TRT] Tactic: 7077887 Time: 0.187187 [TRT] Tactic: 7143423 Time: 0.219218 [TRT] Tactic: 7208959 Time: 0.17974 [TRT] Tactic: 7340031 Time: 0.14013 [TRT] Tactic: 7405567 Time: 0.186797 [TRT] Tactic: 7536639 Time: 0.179609 [TRT] Tactic: 7602175 Time: 0.269376 [TRT] Tactic: 7733247 Time: 0.119193 [TRT] Tactic: 7798783 Time: 0.274036 [TRT] Tactic: 8191999 Time: 0.275495 [TRT] Tactic: 8257535 Time: 0.158749 [TRT] Tactic: 8323071 Time: 0.154427 [TRT] Tactic: 8650751 Time: 0.269791 [TRT] Tactic: 8716287 Time: 0.12362 [TRT] Tactic: 9109503 Time: 0.17237 [TRT] Tactic: 9568255 Time: 0.157552 [TRT] Tactic: 9895935 Time: 0.150339 [TRT] Tactic: 10223615 Time: 0.306354 [TRT] Tactic: 10354687 Time: 0.190651 [TRT] Tactic: 10551295 Time: 0.194193 [TRT] Tactic: 10747903 Time: 0.118281 [TRT] Tactic: 10944511 Time: 0.230521 [TRT] Fastest Tactic: 10747903 Time: 0.118281 [TRT] --------------- Timing Runner: inception_5b/pool_proj + inception_5b/relu_pool_proj (CudnnConvolution) [TRT] Tactic: 0 Time: 0.247344 [TRT] Tactic: 1 Time: 0.208672 [TRT] Tactic: 2 Time: 0.432083 [TRT] Tactic: 4 skipped. Scratch requested: 247709696, available: 33554432 [TRT] Tactic: 5 Time: 7.33378 [TRT] Fastest Tactic: 1 Time: 0.208672 [TRT] --------------- Timing Runner: inception_5b/pool_proj + inception_5b/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5b/pool_proj + inception_5b/relu_pool_proj (CaskConvolution) [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 0.224141 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v0 Tactic: 1698681053543049347 [TRT] Tactic: 1698681053543049347 Time: 0.197734 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 0.179166 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 0.172787 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 0.184063 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v1 Tactic: 5326823351883942011 [TRT] Tactic: 5326823351883942011 Time: 0.174635 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v0 Tactic: 5500448035057547314 [TRT] Tactic: 5500448035057547314 Time: 0.216276 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 0.177943 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 0.212135 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 0.19112 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v1 Tactic: -6576203419454146580 [TRT] Tactic: -6576203419454146580 Time: 0.19638 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 0.221537 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 0.20612 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 0.240391 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 0.217761 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 0.174609 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v0 Tactic: -377491875521947884 [TRT] Tactic: -377491875521947884 Time: 0.182083 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v1 Tactic: -37215280111360163 [TRT] Tactic: -37215280111360163 Time: 0.170521 [TRT] Fastest Tactic: -37215280111360163 Time: 0.170521 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 10747903 [TRT] *************** Autotuning format combination: Float(40768,1,5824,832) -> Float(50176,1,7168,1024) *************** [TRT] --------------- Timing Runner: inception_5b/pool_proj + inception_5b/relu_pool_proj (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5b/pool_proj + inception_5b/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5b/pool_proj + inception_5b/relu_pool_proj (CaskConvolution) [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 3886731678879822788 [TRT] Tactic: 3886731678879822788 Time: 0.147552 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 6629944304117643200 [TRT] Tactic: 6629944304117643200 Time: 0.19487 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 0.195052 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 0.146796 [TRT] Fastest Tactic: -7394439838318485025 Time: 0.146796 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(40768,49,7,1) -> Half(50176,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/pool_proj + inception_5b/relu_pool_proj (CudnnConvolution) [TRT] Tactic: 0 Time: 0.24474 [TRT] Tactic: 1 Time: 0.22565 [TRT] Tactic: 2 Time: 0.421693 [TRT] Tactic: 4 skipped. Scratch requested: 247709696, available: 33554432 [TRT] Tactic: 5 Time: 7.19789 [TRT] Fastest Tactic: 1 Time: 0.22565 [TRT] --------------- Timing Runner: inception_5b/pool_proj + inception_5b/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5b/pool_proj + inception_5b/relu_pool_proj (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 1 [TRT] *************** Autotuning format combination: Half(20384,49:2,7,1) -> Half(50176,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/pool_proj + inception_5b/relu_pool_proj (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] *************** Autotuning format combination: Half(20384,49:2,7,1) -> Half(25088,49:2,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/pool_proj + inception_5b/relu_pool_proj (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 0.150521 [TRT] Tactic: 655359 Time: 0.171172 [TRT] Tactic: 786431 Time: 0.155338 [TRT] Tactic: 851967 Time: 0.103073 [TRT] Tactic: 1179647 Time: 0.084479 [TRT] Tactic: 1310719 Time: 0.286172 [TRT] Tactic: 1376255 Time: 0.158801 [TRT] Tactic: 1441791 Time: 0.117318 [TRT] Tactic: 1507327 Time: 0.10125 [TRT] Tactic: 1638399 Time: 0.143932 [TRT] Tactic: 1835007 Time: 0.142344 [TRT] Tactic: 1900543 Time: 0.17948 [TRT] Tactic: 2097151 Time: 0.115078 [TRT] Tactic: 2162687 Time: 0.171744 [TRT] Tactic: 2293759 Time: 0.167136 [TRT] Tactic: 2359295 Time: 0.103932 [TRT] Tactic: 2686975 Time: 0.242552 [TRT] Tactic: 3080191 Time: 0.0968225 [TRT] Tactic: 3342335 Time: 0.185 [TRT] Tactic: 3407871 Time: 0.092136 [TRT] Tactic: 3538943 Time: 0.0942185 [TRT] Tactic: 3670015 Time: 0.321094 [TRT] Tactic: 3932159 Time: 0.0893495 [TRT] Tactic: 3997695 Time: 0.165911 [TRT] Tactic: 4063231 Time: 0.100625 [TRT] Tactic: 4194303 Time: 0.0815365 [TRT] Tactic: 4259839 Time: 0.115104 [TRT] Tactic: 4325375 Time: 0.123516 [TRT] Tactic: 4521983 Time: 0.164349 [TRT] Tactic: 4587519 Time: 0.121275 [TRT] Tactic: 4653055 Time: 0.102708 [TRT] Tactic: 4915199 Time: 0.0854685 [TRT] Tactic: 4980735 Time: 0.120156 [TRT] Tactic: 5177343 Time: 0.085599 [TRT] Tactic: 5242879 Time: 0.0814065 [TRT] Tactic: 5373951 Time: 0.0848965 [TRT] Tactic: 5439487 Time: 0.120599 [TRT] Tactic: 5570559 Time: 0.095547 [TRT] Tactic: 5636095 Time: 0.102656 [TRT] Tactic: 5701631 Time: 0.150807 [TRT] Tactic: 5767167 Time: 0.107213 [TRT] Tactic: 5832703 Time: 0.0882035 [TRT] Tactic: 5898239 Time: 0.073125 [TRT] Tactic: 6029311 Time: 0.157708 [TRT] Tactic: 6225919 Time: 0.0791925 [TRT] Tactic: 6291455 Time: 0.0837495 [TRT] Tactic: 6422527 Time: 0.090052 [TRT] Tactic: 6750207 Time: 0.105391 [TRT] Tactic: 6815743 Time: 0.0814585 [TRT] Tactic: 6946815 Time: 0.128203 [TRT] Tactic: 7012351 Time: 0.115234 [TRT] Tactic: 7077887 Time: 0.08802 [TRT] Tactic: 7143423 Time: 0.113438 [TRT] Tactic: 7208959 Time: 0.08875 [TRT] Tactic: 7340031 Time: 0.0758855 [TRT] Tactic: 7405567 Time: 0.098568 [TRT] Tactic: 7536639 Time: 0.142292 [TRT] Tactic: 7602175 Time: 0.126146 [TRT] Tactic: 7733247 Time: 0.0696615 [TRT] Tactic: 7798783 Time: 0.154923 [TRT] Tactic: 8191999 Time: 0.133828 [TRT] Tactic: 8257535 Time: 0.084687 [TRT] Tactic: 8323071 Time: 0.119088 [TRT] Tactic: 8650751 Time: 0.127917 [TRT] Tactic: 8716287 Time: 0.0811715 [TRT] Tactic: 9109503 Time: 0.118672 [TRT] Tactic: 9568255 Time: 0.0852605 [TRT] Tactic: 9895935 Time: 0.081797 [TRT] Tactic: 10223615 Time: 0.246224 [TRT] Tactic: 10354687 Time: 0.114479 [TRT] Tactic: 10551295 Time: 0.101432 [TRT] Tactic: 10747903 Time: 0.065156 [TRT] Tactic: 10944511 Time: 0.120078 [TRT] Fastest Tactic: 10747903 Time: 0.065156 [TRT] --------------- Timing Runner: inception_5b/pool_proj + inception_5b/relu_pool_proj (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5b/pool_proj + inception_5b/relu_pool_proj (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: inception_5b/pool_proj + inception_5b/relu_pool_proj (CaskConvolution) [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_interior_nn_v1 Tactic: 3066127711859985668 [TRT] Tactic: 3066127711859985668 Time: 0.0995575 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 0.122838 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 0.104974 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 0.0971355 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] Tactic: 8163473458334948789 Time: 0.091198 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 0.0883595 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 0.091719 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 0.092943 [TRT] inception_5b/pool_proj + inception_5b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_interior_nn_v1 Tactic: -1716393687483585322 [TRT] Tactic: -1716393687483585322 Time: 0.088802 [TRT] Fastest Tactic: -4212163711445252890 Time: 0.0883595 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 10747903 [TRT] *************** Autotuning Reformat:Float(50176,49,7,1) -> Float(50176,1,7168,1024) *************** [TRT] *************** Autotuning Reformat:Float(50176,49,7,1) -> Half(50176,49,7,1) *************** [TRT] *************** Autotuning Reformat:Float(50176,49,7,1) -> Half(25088,49:2,7,1) *************** [TRT] *************** Autotuning Reformat:Float(50176,1,7168,1024) -> Float(50176,49,7,1) *************** [TRT] *************** Autotuning Reformat:Float(50176,1,7168,1024) -> Half(50176,49,7,1) *************** [TRT] *************** Autotuning Reformat:Float(50176,1,7168,1024) -> Half(25088,49:2,7,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,49,7,1) -> Float(50176,49,7,1) *************** [TRT] *************** Autotuning Reformat:Half(50176,49,7,1) -> Float(50176,1,7168,1024) *************** [TRT] *************** Autotuning Reformat:Half(50176,49,7,1) -> Half(25088,49:2,7,1) *************** [TRT] *************** Autotuning Reformat:Half(25088,49:2,7,1) -> Float(50176,49,7,1) *************** [TRT] *************** Autotuning Reformat:Half(25088,49:2,7,1) -> Float(50176,1,7168,1024) *************** [TRT] *************** Autotuning Reformat:Half(25088,49:2,7,1) -> Half(50176,49,7,1) *************** [TRT] *************** Autotuning Reformat:Float(30576,49,7,1) -> Float(50176,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 1.60776 [TRT] Tactic: 0 Time: 0.0078125 [TRT] Fastest Tactic: 0 Time: 0.0078125 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(30576,49,7,1) -> Float(50176,1,7168,1024) *************** [TRT] --------------- Timing Runner: inception_5b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.027708 [TRT] Tactic: 0 Time: 0.0323435 [TRT] Fastest Tactic: 1002 Time: 0.027708 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(30576,49,7,1) -> Half(50176,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.703932 [TRT] Tactic: 0 Time: 0.030026 [TRT] Fastest Tactic: 0 Time: 0.030026 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(30576,49,7,1) -> Half(25088,49:2,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0383595 [TRT] Tactic: 0 Time: 0.0352345 [TRT] Fastest Tactic: 0 Time: 0.0352345 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(30576,1,4368,624) -> Float(50176,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.034896 [TRT] Tactic: 0 Time: 0.0289325 [TRT] Fastest Tactic: 0 Time: 0.0289325 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(30576,1,4368,624) -> Float(50176,1,7168,1024) *************** [TRT] --------------- Timing Runner: inception_5b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.038099 [TRT] Tactic: 0 Time: 0.0311195 [TRT] Fastest Tactic: 0 Time: 0.0311195 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(30576,1,4368,624) -> Half(50176,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0287235 [TRT] Tactic: 0 Time: 0.030208 [TRT] Fastest Tactic: 1002 Time: 0.0287235 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Float(30576,1,4368,624) -> Half(25088,49:2,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.033542 [TRT] Tactic: 0 Time: 0.035729 [TRT] Fastest Tactic: 1002 Time: 0.033542 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(30576,49,7,1) -> Float(50176,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.720286 [TRT] Tactic: 0 Time: 0.030469 [TRT] Fastest Tactic: 0 Time: 0.030469 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(30576,49,7,1) -> Float(50176,1,7168,1024) *************** [TRT] --------------- Timing Runner: inception_5b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0265885 [TRT] Tactic: 0 Time: 0.032682 [TRT] Fastest Tactic: 1002 Time: 0.0265885 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(30576,49,7,1) -> Half(50176,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.716901 [TRT] Tactic: 0 Time: 0.005495 [TRT] Fastest Tactic: 0 Time: 0.005495 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(30576,49,7,1) -> Half(25088,49:2,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.040495 [TRT] Tactic: 0 Time: 0.019375 [TRT] Fastest Tactic: 0 Time: 0.019375 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(15288,49:2,7,1) -> Float(50176,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.035677 [TRT] Tactic: 0 Time: 0.035703 [TRT] Fastest Tactic: 1002 Time: 0.035677 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(15288,49:2,7,1) -> Float(50176,1,7168,1024) *************** [TRT] --------------- Timing Runner: inception_5b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.027604 [TRT] Tactic: 0 Time: 0.0373435 [TRT] Fastest Tactic: 1002 Time: 0.027604 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 1002 [TRT] *************** Autotuning Reformat:Half(15288,49:2,7,1) -> Half(50176,49,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.0451825 [TRT] Tactic: 0 Time: 0.0311195 [TRT] Fastest Tactic: 0 Time: 0.0311195 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Half(15288,49:2,7,1) -> Half(25088,49:2,7,1) *************** [TRT] --------------- Timing Runner: inception_5b/1x1 copy (Reformat) [TRT] Tactic: 1002 Time: 0.037865 [TRT] Tactic: 0 Time: 0.005677 [TRT] Fastest Tactic: 0 Time: 0.005677 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: Reformat Tactic: 0 [TRT] *************** Autotuning Reformat:Float(50176,49,7,1) -> Half(50176,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.54604 [TRT] Tactic: 0 Time: 0.057422 [TRT] Fastest Tactic: 0 Time: 0.057422 [TRT] *************** Autotuning Reformat:Float(50176,49,7,1) -> Half(25088,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0909635 [TRT] Tactic: 0 Time: 0.0453645 [TRT] Fastest Tactic: 0 Time: 0.0453645 [TRT] *************** Autotuning Reformat:Float(50176,1,7168,1024) -> Float(50176,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0915365 [TRT] Tactic: 0 Time: 0.0730995 [TRT] Fastest Tactic: 0 Time: 0.0730995 [TRT] *************** Autotuning Reformat:Float(50176,1,7168,1024) -> Half(50176,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0696355 [TRT] Tactic: 0 Time: 0.073776 [TRT] Fastest Tactic: 1002 Time: 0.0696355 [TRT] *************** Autotuning Reformat:Float(50176,1,7168,1024) -> Half(25088,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0810675 [TRT] Tactic: 0 Time: 0.0872395 [TRT] Fastest Tactic: 1002 Time: 0.0810675 [TRT] *************** Autotuning Reformat:Half(50176,49,7,1) -> Float(50176,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 1.59341 [TRT] Tactic: 0 Time: 0.0487755 [TRT] Fastest Tactic: 0 Time: 0.0487755 [TRT] *************** Autotuning Reformat:Half(50176,49,7,1) -> Half(25088,49:2,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0984115 [TRT] Tactic: 0 Time: 0.044922 [TRT] Fastest Tactic: 0 Time: 0.044922 [TRT] *************** Autotuning Reformat:Half(25088,49:2,7,1) -> Float(50176,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.084948 [TRT] Tactic: 0 Time: 0.039037 [TRT] Fastest Tactic: 0 Time: 0.039037 [TRT] *************** Autotuning Reformat:Half(25088,49:2,7,1) -> Half(50176,49,7,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.107448 [TRT] Tactic: 0 Time: 0.038073 [TRT] Fastest Tactic: 0 Time: 0.038073 [TRT] *************** Autotuning format combination: Float(50176,49,7,1) -> Float(1024,1,1,1) *************** [TRT] --------------- Timing Runner: pool5/7x7_s1 (TiledPooling) [TRT] Tactic: 8192257 Time: 0.107422 [TRT] Tactic: 8257793 Time: 0.084401 [TRT] Tactic: 8323329 Time: 0.0775 [TRT] Tactic: 8388865 Time: 0.0754945 [TRT] Tactic: 8454401 Time: 0.075104 [TRT] Tactic: 8519937 Time: 0.075052 [TRT] Tactic: 8585473 Time: 0.073021 [TRT] Tactic: 8651009 Time: 0.072657 [TRT] Fastest Tactic: 8651009 Time: 0.072657 [TRT] --------------- Timing Runner: pool5/7x7_s1 (CudnnPooling) [TRT] Tactic: -1 Time: 0.0253385 [TRT] Fastest Tactic: -1 Time: 0.0253385 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnPooling Tactic: -1 [TRT] *************** Autotuning format combination: Half(50176,49,7,1) -> Half(1024,1,1,1) *************** [TRT] --------------- Timing Runner: pool5/7x7_s1 (TiledPooling) [TRT] TiledPooling has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: pool5/7x7_s1 (CudnnPooling) [TRT] Tactic: -1 Time: 0.024141 [TRT] Fastest Tactic: -1 Time: 0.024141 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnPooling Tactic: -1 [TRT] *************** Autotuning format combination: Half(25088,49:2,7,1) -> Half(512,1:2,1,1) *************** [TRT] --------------- Timing Runner: pool5/7x7_s1 (TiledPooling) [TRT] Tactic: 8192257 Time: 0.123698 [TRT] Tactic: 8257793 Time: 0.121407 [TRT] Tactic: 8323329 Time: 0.121328 [TRT] Tactic: 8388865 Time: 0.120469 [TRT] Tactic: 8454401 Time: 0.121407 [TRT] Tactic: 8519937 Time: 0.121302 [TRT] Tactic: 8585473 Time: 0.121354 [TRT] Tactic: 8651009 Time: 0.119323 [TRT] Fastest Tactic: 8651009 Time: 0.119323 [TRT] --------------- Timing Runner: pool5/7x7_s1 (CudaPooling) [TRT] Tactic: -3 Time: 0.0217965 [TRT] Fastest Tactic: -3 Time: 0.0217965 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudaPooling Tactic: -3 [TRT] *************** Autotuning Reformat:Float(1024,1,1,1) -> Float(1024,1,1024,1024) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0068485 [TRT] Tactic: 0 Time: 0.005651 [TRT] Fastest Tactic: 0 Time: 0.005651 [TRT] *************** Autotuning Reformat:Float(1024,1,1,1) -> Half(1024,1,1,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.007864 [TRT] Tactic: 0 Time: 0.005521 [TRT] Fastest Tactic: 0 Time: 0.005521 [TRT] *************** Autotuning Reformat:Float(1024,1,1,1) -> Half(512,1:2,1,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0289845 [TRT] Tactic: 0 Time: 0.0046615 [TRT] Fastest Tactic: 0 Time: 0.0046615 [TRT] *************** Autotuning Reformat:Float(1024,1,1024,1024) -> Float(1024,1,1,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.007084 [TRT] Tactic: 0 Time: 0.005573 [TRT] Fastest Tactic: 0 Time: 0.005573 [TRT] *************** Autotuning Reformat:Float(1024,1,1024,1024) -> Half(1024,1,1,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.006198 [TRT] Tactic: 0 Time: 0.005703 [TRT] Fastest Tactic: 0 Time: 0.005703 [TRT] *************** Autotuning Reformat:Float(1024,1,1024,1024) -> Half(512,1:2,1,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0289585 [TRT] Tactic: 0 Time: 0.005573 [TRT] Fastest Tactic: 0 Time: 0.005573 [TRT] *************** Autotuning Reformat:Half(1024,1,1,1) -> Float(1024,1,1,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0079685 [TRT] Tactic: 0 Time: 0.005677 [TRT] Fastest Tactic: 0 Time: 0.005677 [TRT] *************** Autotuning Reformat:Half(1024,1,1,1) -> Float(1024,1,1024,1024) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.007838 [TRT] Tactic: 0 Time: 0.005521 [TRT] Fastest Tactic: 0 Time: 0.005521 [TRT] *************** Autotuning Reformat:Half(1024,1,1,1) -> Half(512,1:2,1,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0302345 [TRT] Tactic: 0 Time: 0.0045575 [TRT] Fastest Tactic: 0 Time: 0.0045575 [TRT] *************** Autotuning Reformat:Half(512,1:2,1,1) -> Float(1024,1,1,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0303385 [TRT] Tactic: 0 Time: 0.003411 [TRT] Fastest Tactic: 0 Time: 0.003411 [TRT] *************** Autotuning Reformat:Half(512,1:2,1,1) -> Float(1024,1,1024,1024) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0292185 [TRT] Tactic: 0 Time: 0.0058855 [TRT] Fastest Tactic: 0 Time: 0.0058855 [TRT] *************** Autotuning Reformat:Half(512,1:2,1,1) -> Half(1024,1,1,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0292185 [TRT] Tactic: 0 Time: 0.004401 [TRT] Fastest Tactic: 0 Time: 0.004401 [TRT] *************** Autotuning format combination: Float(1024,1,1,1) -> Float(1000,1,1,1) *************** [TRT] --------------- Timing Runner: loss3/classifier (CudaDepthwiseConvolution) [TRT] CudaDepthwiseConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: loss3/classifier (FusedConvActConvolution) [TRT] FusedConvActConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: loss3/classifier (CudnnConvolution) [TRT] Tactic: 0 Time: 0.797994 [TRT] Tactic: 1 Time: 0.663828 [TRT] Tactic: 2 Time: 1.17513 [TRT] Tactic: 4 skipped. Scratch requested: 2365751296, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 139539456, available: 33554432 [TRT] Fastest Tactic: 1 Time: 0.663828 [TRT] --------------- Timing Runner: loss3/classifier (CublasConvolution) [TRT] Tactic: 0 Time: 1.63307 [TRT] Tactic: 1 Time: 1.35393 [TRT] Fastest Tactic: 1 Time: 1.35393 [TRT] --------------- Timing Runner: loss3/classifier (CaskConvolution) [TRT] loss3/classifier Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v1 Tactic: 1062367460111450758 [TRT] Tactic: 1062367460111450758 Time: 19.6391 [TRT] loss3/classifier Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v0 Tactic: 1698681053543049347 [TRT] Tactic: 1698681053543049347 Time: 17.8621 [TRT] loss3/classifier Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v1 Tactic: 4501471010995462441 [TRT] Tactic: 4501471010995462441 Time: 3.29935 [TRT] loss3/classifier Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v1 Tactic: 5137655947464784826 [TRT] Tactic: 5137655947464784826 Time: 3.25187 [TRT] loss3/classifier Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v0 Tactic: 5288347012147084929 [TRT] Tactic: 5288347012147084929 Time: 5.42654 [TRT] loss3/classifier Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v1 Tactic: 5326823351883942011 [TRT] Tactic: 5326823351883942011 Time: 5.33664 [TRT] loss3/classifier Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v0 Tactic: 5500448035057547314 [TRT] Tactic: 5500448035057547314 Time: 2.33737 [TRT] loss3/classifier Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v1 Tactic: 6645123197870846056 [TRT] Tactic: 6645123197870846056 Time: 4.08023 [TRT] loss3/classifier Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v0 Tactic: 7144526460361122478 [TRT] Tactic: 7144526460361122478 Time: 4.89479 [TRT] loss3/classifier Set Tactic Name: maxwell_scudnn_128x128_relu_medium_nn_v0 Tactic: -8262349710178828730 [TRT] Tactic: -8262349710178828730 Time: 4.12982 [TRT] loss3/classifier Set Tactic Name: maxwell_scudnn_128x32_relu_interior_nn_v1 Tactic: -6576203419454146580 [TRT] Tactic: -6576203419454146580 Time: 4.43487 [TRT] loss3/classifier Set Tactic Name: maxwell_scudnn_128x32_relu_medium_nn_v0 Tactic: -4787320710726427159 [TRT] Tactic: -4787320710726427159 Time: 4.23253 [TRT] loss3/classifier Set Tactic Name: maxwell_scudnn_128x32_relu_small_nn_v1 Tactic: -3456450830548107839 [TRT] Tactic: -3456450830548107839 Time: 3.78232 [TRT] loss3/classifier Set Tactic Name: maxwell_scudnn_128x64_relu_medium_nn_v0 Tactic: -1218658103698133241 [TRT] Tactic: -1218658103698133241 Time: 3.43513 [TRT] loss3/classifier Set Tactic Name: maxwell_scudnn_128x64_relu_small_nn_v0 Tactic: -836875257600482091 [TRT] Tactic: -836875257600482091 Time: 3.35102 [TRT] loss3/classifier Set Tactic Name: maxwell_scudnn_128x128_relu_small_nn_v1 Tactic: -410470605513481746 [TRT] Tactic: -410470605513481746 Time: 3.23341 [TRT] loss3/classifier Set Tactic Name: maxwell_scudnn_128x128_relu_interior_nn_v0 Tactic: -377491875521947884 [TRT] Tactic: -377491875521947884 Time: 3.23466 [TRT] loss3/classifier Set Tactic Name: maxwell_scudnn_128x64_relu_interior_nn_v1 Tactic: -37215280111360163 [TRT] Tactic: -37215280111360163 Time: 3.16552 [TRT] Fastest Tactic: 5500448035057547314 Time: 2.33737 [TRT] Setting workspace to 139539456enables more tactics for profiling [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnConvolution Tactic: 1 [TRT] *************** Autotuning format combination: Float(1024,1,1024,1024) -> Float(1000,1,1000,1000) *************** [TRT] --------------- Timing Runner: loss3/classifier (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: loss3/classifier (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: loss3/classifier (CaskConvolution) [TRT] loss3/classifier Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 3886731678879822788 [TRT] Tactic: 3886731678879822788 Time: 3.20328 [TRT] loss3/classifier Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_interior_nhwc_tn_v1 Tactic: 6629944304117643200 [TRT] Tactic: 6629944304117643200 Time: 3.37313 [TRT] loss3/classifier Set Tactic Name: maxwell_scudnn_128x32_sliced1x4_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -9153228964338181824 [TRT] Tactic: -9153228964338181824 Time: 3.37576 [TRT] loss3/classifier Set Tactic Name: maxwell_scudnn_128x64_sliced1x2_ldg4_relu_exp_small_nhwc_tn_v1 Tactic: -7394439838318485025 [TRT] Tactic: -7394439838318485025 Time: 2.6674 [TRT] Fastest Tactic: -7394439838318485025 Time: 2.6674 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CaskConvolution Tactic: -7394439838318485025 [TRT] *************** Autotuning format combination: Half(1024,1,1,1) -> Half(1000,1,1,1) *************** [TRT] --------------- Timing Runner: loss3/classifier (CudnnConvolution) [TRT] Tactic: 0 Time: 1.52828 [TRT] Tactic: 1 Time: 1.33674 [TRT] Tactic: 2 Time: 2.31404 [TRT] Tactic: 4 skipped. Scratch requested: 2365751296, available: 33554432 [TRT] Tactic: 5 skipped. Scratch requested: 139539456, available: 33554432 [TRT] Fastest Tactic: 1 Time: 1.33674 [TRT] --------------- Timing Runner: loss3/classifier (CublasConvolution) [TRT] Tactic: 0 Time: 1.63531 [TRT] Tactic: 1 Time: 1.28823 [TRT] Tactic: 4 Time: 1.9874 [TRT] Tactic: 5 Time: 1.20305 [TRT] Fastest Tactic: 5 Time: 1.20305 [TRT] --------------- Timing Runner: loss3/classifier (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] Setting workspace to 139539456enables more tactics for profiling [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CublasConvolution Tactic: 5 [TRT] *************** Autotuning format combination: Half(512,1:2,1,1) -> Half(1000,1,1,1) *************** [TRT] --------------- Timing Runner: loss3/classifier (CaskConvolution) [TRT] CaskConvolution has no valid tactics for this config, skipping [TRT] *************** Autotuning format combination: Half(512,1:2,1,1) -> Half(500,1:2,1,1) *************** [TRT] --------------- Timing Runner: loss3/classifier (FusedConvActConvolution) [TRT] Tactic: 589823 Time: 4.12083 [TRT] Tactic: 655359 Time: 2.24859 [TRT] Tactic: 786431 Time: 3.98943 [TRT] Tactic: 851967 Time: 2.15891 [TRT] Tactic: 1179647 Time: 1.76716 [TRT] Tactic: 1310719 Time: 3.1912 [TRT] Tactic: 1376255 Time: 0.527058 [TRT] Tactic: 1441791 Time: 0.769817 [TRT] Tactic: 1507327 Time: 1.0599 [TRT] Tactic: 1638399 Time: 1.49648 [TRT] Tactic: 1835007 Time: 1.19076 [TRT] Tactic: 1900543 Time: 1.29883 [TRT] Tactic: 2097151 Time: 1.84344 [TRT] Tactic: 2162687 Time: 0.904297 [TRT] Tactic: 2293759 Time: 0.797448 [TRT] Tactic: 2359295 Time: 1.02883 [TRT] Tactic: 2686975 Time: 1.25036 [TRT] Tactic: 3080191 Time: 0.748412 [TRT] Tactic: 3342335 Time: 1.0537 [TRT] Tactic: 3407871 Time: 0.694453 [TRT] Tactic: 3538943 Time: 0.644114 [TRT] Tactic: 3670015 Time: 1.21763 [TRT] Tactic: 3932159 Time: 0.371068 [TRT] Tactic: 3997695 Time: 1.39297 [TRT] Tactic: 4063231 Time: 0.792135 [TRT] Tactic: 4194303 Time: 1.38206 [TRT] Tactic: 4259839 Time: 1.95383 [TRT] Tactic: 4325375 Time: 1.4025 [TRT] Tactic: 4521983 Time: 1.54828 [TRT] Tactic: 4587519 Time: 1.40737 [TRT] Tactic: 4653055 Time: 1.05068 [TRT] Tactic: 4915199 Time: 1.39375 [TRT] Tactic: 4980735 Time: 1.48977 [TRT] Tactic: 5177343 Time: 0.602942 [TRT] Tactic: 5242879 Time: 0.606146 [TRT] Tactic: 5373951 Time: 0.606771 [TRT] Tactic: 5439487 Time: 0.816094 [TRT] Tactic: 5570559 Time: 1.02688 [TRT] Tactic: 5636095 Time: 0.792475 [TRT] Tactic: 5701631 Time: 0.697031 [TRT] Tactic: 5767167 Time: 0.526901 [TRT] Tactic: 5832703 Time: 0.653125 [TRT] Tactic: 5898239 Time: 1.11807 [TRT] Tactic: 6029311 Time: 0.666406 [TRT] Tactic: 6225919 Time: 0.545599 [TRT] Tactic: 6291455 Time: 0.610105 [TRT] Tactic: 6422527 Time: 0.619922 [TRT] Tactic: 6750207 Time: 0.778724 [TRT] Tactic: 6815743 Time: 0.602943 [TRT] Tactic: 6946815 Time: 0.858594 [TRT] Tactic: 7012351 Time: 1.84755 [TRT] Tactic: 7077887 Time: 0.602656 [TRT] Tactic: 7143423 Time: 0.533125 [TRT] Tactic: 7208959 Time: 0.655313 [TRT] Tactic: 7340031 Time: 1.1238 [TRT] Tactic: 7405567 Time: 0.869792 [TRT] Tactic: 7536639 Time: 0.522292 [TRT] Tactic: 7602175 Time: 0.862344 [TRT] Tactic: 7733247 Time: 1.07112 [TRT] Tactic: 7798783 Time: 1.36268 [TRT] Tactic: 8191999 Time: 0.546771 [TRT] Tactic: 8257535 Time: 1.38331 [TRT] Tactic: 8323071 Time: 0.811328 [TRT] Tactic: 8650751 Time: 0.861901 [TRT] Tactic: 8716287 Time: 0.541328 [TRT] Tactic: 9109503 Time: 1.96544 [TRT] Tactic: 9568255 Time: 1.39227 [TRT] Tactic: 9895935 Time: 1.37586 [TRT] Tactic: 10223615 Time: 1.25344 [TRT] Tactic: 10354687 Time: 1.78185 [TRT] Tactic: 10551295 Time: 0.893125 [TRT] Tactic: 10747903 Time: 1.00279 [TRT] Tactic: 10944511 Time: 1.49047 [TRT] Fastest Tactic: 3932159 Time: 0.371068 [TRT] --------------- Timing Runner: loss3/classifier (CudnnConvolution) [TRT] CudnnConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: loss3/classifier (CublasConvolution) [TRT] CublasConvolution has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: loss3/classifier (CaskConvolution) [TRT] loss3/classifier Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_interior_nn_v1 Tactic: 3066127711859985668 [TRT] Tactic: 3066127711859985668 Time: 3.04424 [TRT] loss3/classifier Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_medium_nn_v1 Tactic: 3564772625446233998 [TRT] Tactic: 3564772625446233998 Time: 3.3651 [TRT] loss3/classifier Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_small_nn_v1 Tactic: 5319956359050645452 [TRT] Tactic: 5319956359050645452 Time: 3.18125 [TRT] loss3/classifier Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] Tactic: 7205456024582378848 Time: 2.85893 [TRT] loss3/classifier Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] Tactic: 8163473458334948789 Time: 2.69242 [TRT] loss3/classifier Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] Tactic: -4212163711445252890 Time: 2.71781 [TRT] loss3/classifier Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_medium_nn_v1 Tactic: -3898373634979201110 [TRT] Tactic: -3898373634979201110 Time: 2.78607 [TRT] loss3/classifier Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_small_nn_v1 Tactic: -2409163523992614473 [TRT] Tactic: -2409163523992614473 Time: 2.78016 [TRT] loss3/classifier Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_interior_nn_v1 Tactic: -1716393687483585322 [TRT] Tactic: -1716393687483585322 Time: 2.67229 [TRT] Fastest Tactic: -1716393687483585322 Time: 2.67229 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: FusedConvActConvolution Tactic: 3932159 [TRT] *************** Autotuning Reformat:Float(1000,1,1,1) -> Half(1000,1,1,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.024219 [TRT] Tactic: 0 Time: 0.023594 [TRT] Fastest Tactic: 0 Time: 0.023594 [TRT] *************** Autotuning Reformat:Float(1000,1,1,1) -> Half(500,1:2,1,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.111511 [TRT] Tactic: 0 Time: 0.0131245 [TRT] Fastest Tactic: 0 Time: 0.0131245 [TRT] *************** Autotuning Reformat:Float(1000,1,1000,1000) -> Float(1000,1,1,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.023542 [TRT] Tactic: 0 Time: 0.018333 [TRT] Fastest Tactic: 0 Time: 0.018333 [TRT] *************** Autotuning Reformat:Float(1000,1,1000,1000) -> Half(1000,1,1,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.0236715 [TRT] Tactic: 0 Time: 0.0182815 [TRT] Fastest Tactic: 0 Time: 0.0182815 [TRT] *************** Autotuning Reformat:Float(1000,1,1000,1000) -> Half(500,1:2,1,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.112006 [TRT] Tactic: 0 Time: 0.018411 [TRT] Fastest Tactic: 0 Time: 0.018411 [TRT] *************** Autotuning Reformat:Half(1000,1,1,1) -> Float(1000,1,1,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.028906 [TRT] Tactic: 0 Time: 0.0182815 [TRT] Fastest Tactic: 0 Time: 0.0182815 [TRT] *************** Autotuning Reformat:Half(1000,1,1,1) -> Half(500,1:2,1,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.116953 [TRT] Tactic: 0 Time: 0.013047 [TRT] Fastest Tactic: 0 Time: 0.013047 [TRT] *************** Autotuning Reformat:Half(500,1:2,1,1) -> Float(1000,1,1,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.117058 [TRT] Tactic: 0 Time: 0.0131515 [TRT] Fastest Tactic: 0 Time: 0.0131515 [TRT] *************** Autotuning Reformat:Half(500,1:2,1,1) -> Half(1000,1,1,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.117005 [TRT] Tactic: 0 Time: 0.012969 [TRT] Fastest Tactic: 0 Time: 0.012969 [TRT] *************** Autotuning format combination: Float(1000,1,1,1) -> Float(1000,1,1,1) *************** [TRT] --------------- Timing Runner: prob (CudaSoftMax) [TRT] CudaSoftMax has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: prob (CudnnSoftMax) [TRT] Tactic: 0 Time: 0.044193 [TRT] Fastest Tactic: 0 Time: 0.044193 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnSoftMax Tactic: 0 [TRT] *************** Autotuning format combination: Half(1000,1,1,1) -> Half(1000,1,1,1) *************** [TRT] --------------- Timing Runner: prob (CudaSoftMax) [TRT] CudaSoftMax has no valid tactics for this config, skipping [TRT] --------------- Timing Runner: prob (CudnnSoftMax) [TRT] Tactic: 0 Time: 0.043958 [TRT] Fastest Tactic: 0 Time: 0.043958 [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudnnSoftMax Tactic: 0 [TRT] *************** Autotuning format combination: Half(500,1:2,1,1) -> Half(500,1:2,1,1) *************** [TRT] --------------- Timing Runner: prob (CudaSoftMax) [TRT] Tactic: 18 Time: 0.0312235 [TRT] Fastest Tactic: 18 Time: 0.0312235 [TRT] --------------- Timing Runner: prob (CudnnSoftMax) [TRT] CudnnSoftMax has no valid tactics for this config, skipping [TRT] >>>>>>>>>>>>>>> Chose Runner Type: CudaSoftMax Tactic: 18 [TRT] *************** Autotuning Reformat:Half(1000,1,1,1) -> Float(1000,1,1,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.025677 [TRT] Tactic: 0 Time: 0.018177 [TRT] Fastest Tactic: 0 Time: 0.018177 [TRT] *************** Autotuning Reformat:Half(500,1:2,1,1) -> Float(1000,1,1,1) *************** [TRT] --------------- Timing Runner: Optimizer Reformat (Reformat) [TRT] Tactic: 1002 Time: 0.117526 [TRT] Tactic: 0 Time: 0.013021 [TRT] Fastest Tactic: 0 Time: 0.013021 [TRT] Adding reformat layer: Reformatted Input Tensor 0 to conv1/7x7_s2 + conv1/relu_7x7 (data) from Float(150528,50176,224,1) to Half(100352,50176:2,224,1) [TRT] Adding reformat layer: Reformatted Output Tensor 0 to prob (prob) from Half(500,1:2,1,1) to Float(1000,1,1,1) [TRT] Formats and tactics selection completed in 108.127 seconds. [TRT] After reformat layers: 68 layers [TRT] Block size 33554432 [TRT] Block size 1605632 [TRT] Block size 1204224 [TRT] Block size 401408 [TRT] Block size 401408 [TRT] Total Activation Memory: 37167104 [TRT] Detected 1 inputs and 1 output network tensors. [TRT] conv1/7x7_s2 + conv1/relu_7x7 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_medium_nn_v1 Tactic: 7205456024582378848 [TRT] conv2/3x3_reduce + conv2/relu_3x3_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] conv2/3x3 + conv2/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_winograd_fp16x2_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 4772821744921268633 [TRT] inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] inception_3a/3x3 + inception_3a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_winograd_fp16x2_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 4772821744921268633 [TRT] inception_3a/pool_proj + inception_3a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x32_relu_interior_nn_v1 Tactic: 3066127711859985668 [TRT] inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] inception_3b/3x3 + inception_3b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_winograd_fp16x2_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 4772821744921268633 [TRT] inception_3b/pool_proj + inception_3b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] inception_4a/3x3 + inception_4a/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_winograd_fp16x2_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 4772821744921268633 [TRT] inception_4a/pool_proj + inception_4a/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] inception_4b/3x3 + inception_4b/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_winograd_fp16x2_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 4772821744921268633 [TRT] inception_4b/pool_proj + inception_4b/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] inception_4c/3x3 + inception_4c/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_winograd_fp16x2_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 4772821744921268633 [TRT] inception_4c/pool_proj + inception_4c/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] inception_4d/3x3 + inception_4d/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_winograd_fp16x2_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 4772821744921268633 [TRT] inception_4d/pool_proj + inception_4d/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] inception_4e/3x3 + inception_4e/relu_3x3 Set Tactic Name: maxwell_fp16x2_hcudnn_winograd_fp16x2_128x128_ldg1_ldg4_relu_tile148m_nt_v1 Tactic: 4772821744921268633 [TRT] inception_4e/5x5 + inception_4e/relu_5x5 Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_small_nn_v1 Tactic: -4212163711445252890 [TRT] inception_4e/pool_proj + inception_4e/relu_pool_proj Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x128_relu_interior_nn_v1 Tactic: -1716393687483585322 [TRT] inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce Set Tactic Name: maxwell_fp16x2_hcudnn_fp16x2_128x64_relu_interior_nn_v1 Tactic: 8163473458334948789 [TRT] Layer: Reformatting CopyNode for Input Tensor 0 to conv1/7x7_s2 + conv1/relu_7x7 HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: conv1/7x7_s2 + conv1/relu_7x7 HostPersistent: 2176 DevicePersistent: 100864 [TRT] Layer: pool1/3x3_s2 HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: pool1/norm1 HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: conv2/3x3_reduce + conv2/relu_3x3_reduce HostPersistent: 3200 DevicePersistent: 27648 [TRT] Layer: conv2/3x3 + conv2/relu_3x3 HostPersistent: 512 DevicePersistent: 614912 [TRT] Layer: conv2/norm2 HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: pool2/3x3_s2 HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce HostPersistent: 3200 DevicePersistent: 72704 [TRT] Layer: inception_3a/3x3 + inception_3a/relu_3x3 HostPersistent: 512 DevicePersistent: 614912 [TRT] Layer: inception_3a/5x5 + inception_3a/relu_5x5 HostPersistent: 2192 DevicePersistent: 0 [TRT] Layer: inception_3a/pool HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: inception_3a/pool_proj + inception_3a/relu_pool_proj HostPersistent: 3200 DevicePersistent: 17408 [TRT] Layer: inception_3a/1x1 copy HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce HostPersistent: 3200 DevicePersistent: 153088 [TRT] Layer: inception_3b/3x3 + inception_3b/relu_3x3 HostPersistent: 512 DevicePersistent: 1229312 [TRT] Layer: inception_3b/5x5 + inception_3b/relu_5x5 HostPersistent: 2192 DevicePersistent: 0 [TRT] Layer: inception_3b/pool HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: inception_3b/pool_proj + inception_3b/relu_pool_proj HostPersistent: 3200 DevicePersistent: 37888 [TRT] Layer: inception_3b/1x1 copy HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: pool3/3x3_s2 HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce HostPersistent: 3200 DevicePersistent: 293888 [TRT] Layer: inception_4a/3x3 + inception_4a/relu_3x3 HostPersistent: 512 DevicePersistent: 1048064 [TRT] Layer: inception_4a/5x5 + inception_4a/relu_5x5 HostPersistent: 2192 DevicePersistent: 0 [TRT] Layer: inception_4a/pool HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: inception_4a/pool_proj + inception_4a/relu_pool_proj HostPersistent: 3200 DevicePersistent: 62976 [TRT] Layer: inception_4a/1x1 copy HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce HostPersistent: 3200 DevicePersistent: 305152 [TRT] Layer: inception_4b/3x3 + inception_4b/relu_3x3 HostPersistent: 512 DevicePersistent: 1254912 [TRT] Layer: inception_4b/5x5 + inception_4b/relu_5x5 HostPersistent: 2192 DevicePersistent: 0 [TRT] Layer: inception_4b/pool HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: inception_4b/pool_proj + inception_4b/relu_pool_proj HostPersistent: 3200 DevicePersistent: 67072 [TRT] Layer: inception_4b/1x1 copy HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce HostPersistent: 3200 DevicePersistent: 288768 [TRT] Layer: inception_4c/3x3 + inception_4c/relu_3x3 HostPersistent: 512 DevicePersistent: 1638912 [TRT] Layer: inception_4c/5x5 + inception_4c/relu_5x5 HostPersistent: 2192 DevicePersistent: 0 [TRT] Layer: inception_4c/pool HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: inception_4c/pool_proj + inception_4c/relu_pool_proj HostPersistent: 3200 DevicePersistent: 67072 [TRT] Layer: inception_4c/1x1 copy HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce HostPersistent: 3200 DevicePersistent: 296960 [TRT] Layer: inception_4d/3x3 + inception_4d/relu_3x3 HostPersistent: 512 DevicePersistent: 2074624 [TRT] Layer: inception_4d/5x5 + inception_4d/relu_5x5 HostPersistent: 2192 DevicePersistent: 0 [TRT] Layer: inception_4d/pool HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: inception_4d/pool_proj + inception_4d/relu_pool_proj HostPersistent: 3200 DevicePersistent: 67072 [TRT] Layer: inception_4d/1x1 copy HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce HostPersistent: 3200 DevicePersistent: 475648 [TRT] Layer: inception_4e/3x3 + inception_4e/relu_3x3 HostPersistent: 512 DevicePersistent: 2561024 [TRT] Layer: inception_4e/5x5 + inception_4e/relu_5x5 HostPersistent: 1664 DevicePersistent: 206336 [TRT] Layer: inception_4e/pool HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: inception_4e/pool_proj + inception_4e/relu_pool_proj HostPersistent: 3200 DevicePersistent: 136704 [TRT] Layer: inception_4e/1x1 copy HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: pool4/3x3_s2 HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce HostPersistent: 2192 DevicePersistent: 0 [TRT] Layer: inception_5a/3x3 + inception_5a/relu_3x3 HostPersistent: 2192 DevicePersistent: 0 [TRT] Layer: inception_5a/5x5 + inception_5a/relu_5x5 HostPersistent: 2192 DevicePersistent: 0 [TRT] Layer: inception_5a/pool HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: inception_5a/pool_proj + inception_5a/relu_pool_proj HostPersistent: 2192 DevicePersistent: 0 [TRT] Layer: inception_5a/1x1 copy HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce HostPersistent: 3200 DevicePersistent: 1040384 [TRT] Layer: inception_5b/3x3 + inception_5b/relu_3x3 HostPersistent: 2192 DevicePersistent: 0 [TRT] Layer: inception_5b/5x5 + inception_5b/relu_5x5 HostPersistent: 2192 DevicePersistent: 0 [TRT] Layer: inception_5b/pool HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: inception_5b/pool_proj + inception_5b/relu_pool_proj HostPersistent: 2192 DevicePersistent: 0 [TRT] Layer: inception_5b/1x1 copy HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: pool5/7x7_s1 HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: loss3/classifier HostPersistent: 2192 DevicePersistent: 0 [TRT] Layer: prob HostPersistent: 0 DevicePersistent: 0 [TRT] Layer: Reformatting CopyNode for Output Tensor 0 to prob HostPersistent: 0 DevicePersistent: 0 [TRT] Total Host Persistent Memory: 89824 [TRT] Total Device Persistent Memory: 14754304 [TRT] Total Scratch Memory: 0 [TRT] [MemUsageStats] Peak memory usage of TRT CPU/GPU memory allocators: CPU 15 MiB, GPU 48 MiB [TRT] Using cublas a tactic source [TRT] [MemUsageChange] Init cuBLAS/cuBLASLt: CPU +0, GPU +0, now: CPU 942, GPU 3463 (MiB) [TRT] Using cuDNN as a tactic source [TRT] [MemUsageChange] Init cuDNN: CPU +0, GPU +0, now: CPU 942, GPU 3463 (MiB) [TRT] [MemUsageChange] Init cuBLAS/cuBLASLt: CPU +0, GPU +0, now: CPU 942, GPU 3463 (MiB) [TRT] Engine generation completed in 114.58 seconds. [TRT] Deleting timing cache: 913 entries, 379 hits [TRT] [MemUsageChange] Init cuBLAS/cuBLASLt: CPU +0, GPU +0, now: CPU 942, GPU 3463 (MiB) [TRT] Engine Layer Information: Layer(Reformat): Reformatting CopyNode for Input Tensor 0 to conv1/7x7_s2 + conv1/relu_7x7, Tactic: 0, data[Float(-2,3,224,224)] -> Reformatted Input Tensor 0 to conv1/7x7_s2 + conv1/relu_7x7[Half(-2,3,224,224)] Layer(CaskConvolution): conv1/7x7_s2 + conv1/relu_7x7, Tactic: 7205456024582378848, Reformatted Input Tensor 0 to conv1/7x7_s2 + conv1/relu_7x7[Half(-2,3,224,224)] -> conv1/7x7_s2[Half(-2,64,112,112)] Layer(TiledPooling): pool1/3x3_s2, Tactic: 6947073, conv1/7x7_s2[Half(-2,64,112,112)] -> pool1/3x3_s2[Half(-2,64,56,56)] Layer(CudaLRN): pool1/norm1, Tactic: 0, pool1/3x3_s2[Half(-2,64,56,56)] -> pool1/norm1[Half(-2,64,56,56)] Layer(CaskConvolution): conv2/3x3_reduce + conv2/relu_3x3_reduce, Tactic: 8163473458334948789, pool1/norm1[Half(-2,64,56,56)] -> conv2/3x3_reduce[Half(-2,64,56,56)] Layer(CaskConvolution): conv2/3x3 + conv2/relu_3x3, Tactic: 4772821744921268633, conv2/3x3_reduce[Half(-2,64,56,56)] -> conv2/3x3[Half(-2,192,56,56)] Layer(CudaLRN): conv2/norm2, Tactic: 0, conv2/3x3[Half(-2,192,56,56)] -> conv2/norm2[Half(-2,192,56,56)] Layer(TiledPooling): pool2/3x3_s2, Tactic: 2621697, conv2/norm2[Half(-2,192,56,56)] -> pool2/3x3_s2[Half(-2,192,28,28)] Layer(CaskConvolution): inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce, Tactic: 8163473458334948789, pool2/3x3_s2[Half(-2,192,28,28)] -> inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce[Half(-2,176,28,28)] Layer(CaskConvolution): inception_3a/3x3 + inception_3a/relu_3x3, Tactic: 4772821744921268633, inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce[Half(-2,96,28,28)] -> inception_3a/output[Half(-2,128,28,28)] Layer(FusedConvActConvolution): inception_3a/5x5 + inception_3a/relu_5x5, Tactic: 6553599, inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce[Half(-2,16,28,28)] -> inception_3a/output[Half(-2,32,28,28)] Layer(TiledPooling): inception_3a/pool, Tactic: 6553857, pool2/3x3_s2[Half(-2,192,28,28)] -> inception_3a/pool[Half(-2,192,28,28)] Layer(CaskConvolution): inception_3a/pool_proj + inception_3a/relu_pool_proj, Tactic: 3066127711859985668, inception_3a/pool[Half(-2,192,28,28)] -> inception_3a/output[Half(-2,32,28,28)] Layer(Reformat): inception_3a/1x1 copy, Tactic: 0, inception_3a/1x1 + inception_3a/relu_1x1 || inception_3a/3x3_reduce + inception_3a/relu_3x3_reduce || inception_3a/5x5_reduce + inception_3a/relu_5x5_reduce[Half(-2,64,28,28)] -> inception_3a/output[Half(-2,64,28,28)] Layer(CaskConvolution): inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce, Tactic: 8163473458334948789, inception_3a/output[Half(-2,256,28,28)] -> inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce[Half(-2,288,28,28)] Layer(CaskConvolution): inception_3b/3x3 + inception_3b/relu_3x3, Tactic: 4772821744921268633, inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce[Half(-2,128,28,28)] -> inception_3b/output[Half(-2,192,28,28)] Layer(FusedConvActConvolution): inception_3b/5x5 + inception_3b/relu_5x5, Tactic: 393215, inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce[Half(-2,32,28,28)] -> inception_3b/output[Half(-2,96,28,28)] Layer(TiledPooling): inception_3b/pool, Tactic: 6553857, inception_3a/output[Half(-2,256,28,28)] -> inception_3b/pool[Half(-2,256,28,28)] Layer(CaskConvolution): inception_3b/pool_proj + inception_3b/relu_pool_proj, Tactic: 8163473458334948789, inception_3b/pool[Half(-2,256,28,28)] -> inception_3b/output[Half(-2,64,28,28)] Layer(Reformat): inception_3b/1x1 copy, Tactic: 0, inception_3b/1x1 + inception_3b/relu_1x1 || inception_3b/3x3_reduce + inception_3b/relu_3x3_reduce || inception_3b/5x5_reduce + inception_3b/relu_5x5_reduce[Half(-2,128,28,28)] -> inception_3b/output[Half(-2,128,28,28)] Layer(TiledPooling): pool3/3x3_s2, Tactic: 2359553, inception_3b/output[Half(-2,480,28,28)] -> pool3/3x3_s2[Half(-2,480,14,14)] Layer(CaskConvolution): inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce, Tactic: 8163473458334948789, pool3/3x3_s2[Half(-2,480,14,14)] -> inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce[Half(-2,304,14,14)] Layer(CaskConvolution): inception_4a/3x3 + inception_4a/relu_3x3, Tactic: 4772821744921268633, inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce[Half(-2,96,14,14)] -> inception_4a/output[Half(-2,208,14,14)] Layer(FusedConvActConvolution): inception_4a/5x5 + inception_4a/relu_5x5, Tactic: 9240575, inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce[Half(-2,16,14,14)] -> inception_4a/output[Half(-2,48,14,14)] Layer(TiledPooling): inception_4a/pool, Tactic: 5177601, pool3/3x3_s2[Half(-2,480,14,14)] -> inception_4a/pool[Half(-2,480,14,14)] Layer(CaskConvolution): inception_4a/pool_proj + inception_4a/relu_pool_proj, Tactic: 8163473458334948789, inception_4a/pool[Half(-2,480,14,14)] -> inception_4a/output[Half(-2,64,14,14)] Layer(Reformat): inception_4a/1x1 copy, Tactic: 0, inception_4a/1x1 + inception_4a/relu_1x1 || inception_4a/3x3_reduce + inception_4a/relu_3x3_reduce || inception_4a/5x5_reduce + inception_4a/relu_5x5_reduce[Half(-2,192,14,14)] -> inception_4a/output[Half(-2,192,14,14)] Layer(CaskConvolution): inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce, Tactic: 8163473458334948789, inception_4a/output[Half(-2,512,14,14)] -> inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce[Half(-2,296,14,14)] Layer(CaskConvolution): inception_4b/3x3 + inception_4b/relu_3x3, Tactic: 4772821744921268633, inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce[Half(-2,112,14,14)] -> inception_4b/output[Half(-2,224,14,14)] Layer(FusedConvActConvolution): inception_4b/5x5 + inception_4b/relu_5x5, Tactic: 9240575, inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce[Half(-2,24,14,14)] -> inception_4b/output[Half(-2,64,14,14)] Layer(TiledPooling): inception_4b/pool, Tactic: 4260097, inception_4a/output[Half(-2,512,14,14)] -> inception_4b/pool[Half(-2,512,14,14)] Layer(CaskConvolution): inception_4b/pool_proj + inception_4b/relu_pool_proj, Tactic: 8163473458334948789, inception_4b/pool[Half(-2,512,14,14)] -> inception_4b/output[Half(-2,64,14,14)] Layer(Reformat): inception_4b/1x1 copy, Tactic: 0, inception_4b/1x1 + inception_4b/relu_1x1 || inception_4b/3x3_reduce + inception_4b/relu_3x3_reduce || inception_4b/5x5_reduce + inception_4b/relu_5x5_reduce[Half(-2,160,14,14)] -> inception_4b/output[Half(-2,160,14,14)] Layer(CaskConvolution): inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce, Tactic: 8163473458334948789, inception_4b/output[Half(-2,512,14,14)] -> inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce[Half(-2,280,14,14)] Layer(CaskConvolution): inception_4c/3x3 + inception_4c/relu_3x3, Tactic: 4772821744921268633, inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce[Half(-2,128,14,14)] -> inception_4c/output[Half(-2,256,14,14)] Layer(FusedConvActConvolution): inception_4c/5x5 + inception_4c/relu_5x5, Tactic: 9240575, inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce[Half(-2,24,14,14)] -> inception_4c/output[Half(-2,64,14,14)] Layer(TiledPooling): inception_4c/pool, Tactic: 4260097, inception_4b/output[Half(-2,512,14,14)] -> inception_4c/pool[Half(-2,512,14,14)] Layer(CaskConvolution): inception_4c/pool_proj + inception_4c/relu_pool_proj, Tactic: 8163473458334948789, inception_4c/pool[Half(-2,512,14,14)] -> inception_4c/output[Half(-2,64,14,14)] Layer(Reformat): inception_4c/1x1 copy, Tactic: 0, inception_4c/1x1 + inception_4c/relu_1x1 || inception_4c/3x3_reduce + inception_4c/relu_3x3_reduce || inception_4c/5x5_reduce + inception_4c/relu_5x5_reduce[Half(-2,128,14,14)] -> inception_4c/output[Half(-2,128,14,14)] Layer(CaskConvolution): inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce, Tactic: 8163473458334948789, inception_4c/output[Half(-2,512,14,14)] -> inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce[Half(-2,288,14,14)] Layer(CaskConvolution): inception_4d/3x3 + inception_4d/relu_3x3, Tactic: 4772821744921268633, inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce[Half(-2,144,14,14)] -> inception_4d/output[Half(-2,288,14,14)] Layer(FusedConvActConvolution): inception_4d/5x5 + inception_4d/relu_5x5, Tactic: 9240575, inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce[Half(-2,32,14,14)] -> inception_4d/output[Half(-2,64,14,14)] Layer(TiledPooling): inception_4d/pool, Tactic: 4260097, inception_4c/output[Half(-2,512,14,14)] -> inception_4d/pool[Half(-2,512,14,14)] Layer(CaskConvolution): inception_4d/pool_proj + inception_4d/relu_pool_proj, Tactic: 8163473458334948789, inception_4d/pool[Half(-2,512,14,14)] -> inception_4d/output[Half(-2,64,14,14)] Layer(Reformat): inception_4d/1x1 copy, Tactic: 0, inception_4d/5x5_reduce + inception_4d/relu_5x5_reduce || inception_4d/1x1 + inception_4d/relu_1x1 || inception_4d/3x3_reduce + inception_4d/relu_3x3_reduce[Half(-2,112,14,14)] -> inception_4d/output[Half(-2,112,14,14)] Layer(CaskConvolution): inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce, Tactic: 8163473458334948789, inception_4d/output[Half(-2,528,14,14)] -> inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce[Half(-2,448,14,14)] Layer(CaskConvolution): inception_4e/3x3 + inception_4e/relu_3x3, Tactic: 4772821744921268633, inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce[Half(-2,160,14,14)] -> inception_4e/output[Half(-2,320,14,14)] Layer(CaskConvolution): inception_4e/5x5 + inception_4e/relu_5x5, Tactic: -4212163711445252890, inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce[Half(-2,32,14,14)] -> inception_4e/output[Half(-2,128,14,14)] Layer(TiledPooling): inception_4e/pool, Tactic: 5177601, inception_4d/output[Half(-2,528,14,14)] -> inception_4e/pool[Half(-2,528,14,14)] Layer(CaskConvolution): inception_4e/pool_proj + inception_4e/relu_pool_proj, Tactic: -1716393687483585322, inception_4e/pool[Half(-2,528,14,14)] -> inception_4e/output[Half(-2,128,14,14)] Layer(Reformat): inception_4e/1x1 copy, Tactic: 0, inception_4e/1x1 + inception_4e/relu_1x1 || inception_4e/3x3_reduce + inception_4e/relu_3x3_reduce || inception_4e/5x5_reduce + inception_4e/relu_5x5_reduce[Half(-2,256,14,14)] -> inception_4e/output[Half(-2,256,14,14)] Layer(TiledPooling): pool4/3x3_s2, Tactic: 2359553, inception_4e/output[Half(-2,832,14,14)] -> pool4/3x3_s2[Half(-2,832,7,7)] Layer(FusedConvActConvolution): inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce, Tactic: 10747903, pool4/3x3_s2[Half(-2,832,7,7)] -> inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce[Half(-2,448,7,7)] Layer(FusedConvActConvolution): inception_5a/3x3 + inception_5a/relu_3x3, Tactic: 9043967, inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce[Half(-2,160,7,7)] -> inception_5a/output[Half(-2,320,7,7)] Layer(FusedConvActConvolution): inception_5a/5x5 + inception_5a/relu_5x5, Tactic: 7929855, inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce[Half(-2,32,7,7)] -> inception_5a/output[Half(-2,128,7,7)] Layer(TiledPooling): inception_5a/pool, Tactic: 5112065, pool4/3x3_s2[Half(-2,832,7,7)] -> inception_5a/pool[Half(-2,832,7,7)] Layer(FusedConvActConvolution): inception_5a/pool_proj + inception_5a/relu_pool_proj, Tactic: 10747903, inception_5a/pool[Half(-2,832,7,7)] -> inception_5a/output[Half(-2,128,7,7)] Layer(Reformat): inception_5a/1x1 copy, Tactic: 0, inception_5a/1x1 + inception_5a/relu_1x1 || inception_5a/3x3_reduce + inception_5a/relu_3x3_reduce || inception_5a/5x5_reduce + inception_5a/relu_5x5_reduce[Half(-2,256,7,7)] -> inception_5a/output[Half(-2,256,7,7)] Layer(CaskConvolution): inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce, Tactic: 8163473458334948789, inception_5a/output[Half(-2,832,7,7)] -> inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce[Half(-2,624,7,7)] Layer(FusedConvActConvolution): inception_5b/3x3 + inception_5b/relu_3x3, Tactic: 9043967, inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce[Half(-2,192,7,7)] -> inception_5b/output[Half(-2,384,7,7)] Layer(FusedConvActConvolution): inception_5b/5x5 + inception_5b/relu_5x5, Tactic: 9371647, inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce[Half(-2,48,7,7)] -> inception_5b/output[Half(-2,128,7,7)] Layer(TiledPooling): inception_5b/pool, Tactic: 5112065, inception_5a/output[Half(-2,832,7,7)] -> inception_5b/pool[Half(-2,832,7,7)] Layer(FusedConvActConvolution): inception_5b/pool_proj + inception_5b/relu_pool_proj, Tactic: 10747903, inception_5b/pool[Half(-2,832,7,7)] -> inception_5b/output[Half(-2,128,7,7)] Layer(Reformat): inception_5b/1x1 copy, Tactic: 0, inception_5b/1x1 + inception_5b/relu_1x1 || inception_5b/3x3_reduce + inception_5b/relu_3x3_reduce || inception_5b/5x5_reduce + inception_5b/relu_5x5_reduce[Half(-2,384,7,7)] -> inception_5b/output[Half(-2,384,7,7)] Layer(CudaPooling): pool5/7x7_s1, Tactic: -3, inception_5b/output[Half(-2,1024,7,7)] -> pool5/7x7_s1[Half(-2,1024,1,1)] Layer(FusedConvActConvolution): loss3/classifier, Tactic: 3932159, pool5/7x7_s1[Half(-2,1024,1,1)] -> loss3/classifier[Half(-2,1000,1,1)] Layer(CudaSoftMax): prob, Tactic: 18, loss3/classifier[Half(-2,1000,1,1)] -> Reformatted Output Tensor 0 to prob[Half(-2,1000,1,1)] Layer(Reformat): Reformatting CopyNode for Output Tensor 0 to prob, Tactic: 0, Reformatted Output Tensor 0 to prob[Half(-2,1000,1,1)] -> prob[Float(-2,1000,1,1)] [TRT] [MemUsageSnapshot] Builder end: CPU 942 MiB, GPU 3463 MiB [TRT] device GPU, completed building CUDA engine [TRT] network profiling complete, writing engine cache to networks/bvlc_googlenet.caffemodel.1.1.8001.GPU.FP16.engine [TRT] device GPU, completed writing engine cache to networks/bvlc_googlenet.caffemodel.1.1.8001.GPU.FP16.engine [TRT] device GPU, loaded networks/bvlc_googlenet.caffemodel [TRT] [MemUsageChange] Init CUDA: CPU +0, GPU +0, now: CPU 974, GPU 3478 (MiB) [TRT] Loaded engine size: 20 MB [TRT] [MemUsageSnapshot] deserializeCudaEngine begin: CPU 974 MiB, GPU 3478 MiB [TRT] Using cublas a tactic source [TRT] [MemUsageChange] Init cuBLAS/cuBLASLt: CPU +0, GPU +0, now: CPU 975, GPU 3478 (MiB) [TRT] Using cuDNN as a tactic source [TRT] [MemUsageChange] Init cuDNN: CPU +0, GPU +0, now: CPU 975, GPU 3478 (MiB) [TRT] [MemUsageChange] Init cuBLAS/cuBLASLt: CPU +0, GPU +0, now: CPU 975, GPU 3478 (MiB) [TRT] Deserialization required 151634 microseconds. [TRT] [MemUsageSnapshot] deserializeCudaEngine end: CPU 975 MiB, GPU 3478 MiB [TRT] [MemUsageSnapshot] ExecutionContext creation begin: CPU 975 MiB, GPU 3478 MiB [TRT] Using cublas a tactic source [TRT] [MemUsageChange] Init cuBLAS/cuBLASLt: CPU +0, GPU +0, now: CPU 975, GPU 3478 (MiB) [TRT] Using cuDNN as a tactic source [TRT] [MemUsageChange] Init cuDNN: CPU +0, GPU +0, now: CPU 975, GPU 3478 (MiB) [TRT] Total per-runner device memory is 14754304 [TRT] Total per-runner host memory is 89824 [TRT] Allocated activation device memory of size 3612672 [TRT] [MemUsageSnapshot] ExecutionContext creation end: CPU 975 MiB, GPU 3479 MiB [TRT] [TRT] CUDA engine context initialized on device GPU: [TRT] -- layers 68 [TRT] -- maxBatchSize 1 [TRT] -- deviceMemory 3612672 [TRT] -- bindings 2 [TRT] binding 0 -- index 0 -- name 'data' -- type FP32 -- in/out INPUT -- # dims 3 -- dim #0 3 -- dim #1 224 -- dim #2 224 [TRT] binding 1 -- index 1 -- name 'prob' -- type FP32 -- in/out OUTPUT -- # dims 3 -- dim #0 1000 -- dim #1 1 -- dim #2 1 [TRT] [TRT] binding to input 0 data binding index: 0 [TRT] binding to input 0 data dims (b=1 c=3 h=224 w=224) size=602112 [TRT] binding to output 0 prob binding index: 1 [TRT] binding to output 0 prob dims (b=1 c=1000 h=1 w=1) size=4000 [TRT] [TRT] device GPU, networks/bvlc_googlenet.caffemodel initialized. [TRT] imageNet -- loaded 1000 class info entries [TRT] imageNet -- networks/bvlc_googlenet.caffemodel initialized. [video] created imageLoader from file:///jetson-inference/build/aarch64/bin/images/orange_0.jpg ------------------------------------------------ imageLoader video options: ------------------------------------------------ -- URI: file:///jetson-inference/build/aarch64/bin/images/orange_0.jpg - protocol: file - location: images/orange_0.jpg - extension: jpg -- deviceType: file -- ioType: input -- codec: unknown -- width: 0 -- height: 0 -- frameRate: 0.000000 -- bitRate: 0 -- numBuffers: 4 -- zeroCopy: true -- flipMethod: none -- loop: 0 -- rtspLatency 2000 ------------------------------------------------ [video] created imageWriter from file:///jetson-inference/build/aarch64/bin/images/test/output_0.jpg ------------------------------------------------ imageWriter video options: ------------------------------------------------ -- URI: file:///jetson-inference/build/aarch64/bin/images/test/output_0.jpg - protocol: file - location: images/test/output_0.jpg - extension: jpg -- deviceType: file -- ioType: output -- codec: unknown -- width: 0 -- height: 0 -- frameRate: 0.000000 -- bitRate: 0 -- numBuffers: 4 -- zeroCopy: true -- flipMethod: none -- loop: 0 -- rtspLatency 2000 ------------------------------------------------ [OpenGL] failed to open X11 server connection. [OpenGL] failed to create X11 Window. [image] loaded 'images/orange_0.jpg' (1024x683, 3 channels) Traceback (most recent call last): File "./imagenet.py", line 68, in class_id, confidence = net.Classify(img) Exception: jetson.inference -- imageNet.Classify() encountered an error classifying the image root@jetson:/jetson-inference/build/aarch64/bin# exit exit jetson@jetson:~/jetson-inference$ exit logout