&&&& RUNNING TensorRT.sample_onnx_mnist [TensorRT v8200] # D:\Lib\TensorRT-8.2.0.6\bin\sample_onnx_mnist.exe
[11/20/2021-16:47:05] [I] Building and running a GPU inference engine for Onnx MNIST
[11/20/2021-16:47:06] [I] [TRT] [MemUsageChange] Init CUDA: CPU +739, GPU +0, now: CPU 9213, GPU 1284 (MiB)
[11/20/2021-16:47:06] [I] [TRT] ----------------------------------------------------------------
[11/20/2021-16:47:06] [I] [TRT] Input filename:   D:\Python\yolov4-pytorch\\yolov4_author_batch1_180.onnx
[11/20/2021-16:47:06] [I] [TRT] ONNX IR version:  0.0.6
[11/20/2021-16:47:06] [I] [TRT] Opset version:    11
[11/20/2021-16:47:06] [I] [TRT] Producer name:    pytorch
[11/20/2021-16:47:06] [I] [TRT] Producer version: 1.8
[11/20/2021-16:47:06] [I] [TRT] Domain:
[11/20/2021-16:47:06] [I] [TRT] Model version:    0
[11/20/2021-16:47:06] [I] [TRT] Doc string:
[11/20/2021-16:47:06] [I] [TRT] ----------------------------------------------------------------
Run in FP16 mode.
Start buildEngineWithConfig
[11/20/2021-16:47:09] [I] [TRT] [MemUsageSnapshot] Builder begin: CPU 9607 MiB, GPU 1284 MiB
[11/20/2021-16:47:10] [W] [TRT] TensorRT was linked against cuBLAS/cuBLAS LT 11.6.1 but loaded cuBLAS/cuBLAS LT 11.3.0
[11/20/2021-16:47:10] [I] [TRT] [MemUsageChange] Init cuBLAS/cuBLASLt: CPU +742, GPU +266, now: CPU 10356, GPU 1550 (MiB)
[11/20/2021-16:47:11] [I] [TRT] [MemUsageChange] Init cuDNN: CPU +497, GPU +258, now: CPU 10853, GPU 1808 (MiB)
[11/20/2021-16:47:11] [W] [TRT] TensorRT was linked against cuDNN 8.2.1 but loaded cuDNN 8.0.5
[11/20/2021-16:47:11] [I] [TRT] Local timing cache in use. Profiling results in this builder pass will not be stored.
[11/20/2021-16:49:26] [I] [TRT] Some tactics do not have sufficient workspace memory to run. Increasing workspace size may increase performance, please check verbose output.
[11/20/2021-17:29:04] [I] [TRT] [BlockAssignment] Algorithm Linear took 0.0088ms to assign 161 blocks to 161 nodes requiring 255066136 bytes.
[11/20/2021-17:29:04] [I] [TRT] Total Activation Memory: 255066136
[11/20/2021-17:29:04] [I] [TRT] Detected 1 inputs and 3 output network tensors.
[11/20/2021-17:29:05] [I] [TRT] Total Host Persistent Memory: 255296
[11/20/2021-17:29:05] [I] [TRT] Total Device Persistent Memory: 131863552
[11/20/2021-17:29:05] [I] [TRT] Total Scratch Memory: 0
[11/20/2021-17:29:05] [I] [TRT] [MemUsageStats] Peak memory usage of TRT CPU/GPU memory allocators: CPU 137 MiB, GPU 16 MiB
[11/20/2021-17:29:05] [I] [TRT] [BlockAssignment] Algorithm ShiftNTopDown took 34.3711ms to assign 6 blocks to 160 nodes requiring 51118080 bytes.
[11/20/2021-17:29:05] [W] [TRT] TensorRT was linked against cuBLAS/cuBLAS LT 11.6.1 but loaded cuBLAS/cuBLAS LT 11.3.0
[11/20/2021-17:29:05] [I] [TRT] [MemUsageChange] Init cuBLAS/cuBLASLt: CPU +1, GPU +8, now: CPU 12310, GPU 2164 (MiB)
[11/20/2021-17:29:05] [I] [TRT] [MemUsageChange] Init cuDNN: CPU +5, GPU +8, now: CPU 12315, GPU 2172 (MiB)
[11/20/2021-17:29:05] [W] [TRT] TensorRT was linked against cuDNN 8.2.1 but loaded cuDNN 8.0.5
[11/20/2021-17:29:05] [I] [TRT] [MemUsageSnapshot] Builder end: CPU 12313 MiB, GPU 2138 MiB
生成engine文件时间:2.51581e+06ms
Finish buildEngineWithConfig