NVIDIA-AI-IOT / torch2trt

An easy to use PyTorch to TensorRT converter

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

repvggblock in torch2trt

chennaikai11 opened this issue · comments

I encountered an issue when using torch2trt to convert my YOLOX model. It worked fine originally, but when I replaced the original CSPBlock with RepVGGBlock, torch2trt failed to convert the model. There were no error messages, but the conversion process was terminated prematurely. Here is the output:
(yolox) R:\YOLOX>python tool/trt.py --onlypth
2023-12-07 20:53:47.909 | INFO | main:main:63 - loaded checkpoint done.
[12/07/2023-20:53:50] [TRT] [I] [MemUsageChange] Init CUDA: CPU +776, GPU +0, now: CPU 13024, GPU 2274 (MiB)
(yolox) R:\YOLOX>
I suspect that this issue might be due to some operations in RepVGGBlock not being supported by torch2trt. Could you please provide some guidance on how to resolve this issue? Thank you.