2

https://github.com/ultralytics/yolov3/

When I started training yolov3 with 'python train.py' this happened, what is the reason?

Traceback (most recent call last):
    File "train.py", line 415, in <module>
        train()  # train normally
    File "train.py", line 279, in train
        scaled_loss.backward()
    File "C:\Users\zst\anaconda3\lib\site-packages\torch\tensor.py", line 198, in backward
        torch.autograd.backward(self, gradient, retain_graph, create_graph)
    File "C:\Users\zst\anaconda3\lib\site-packages\torch\autograd\__init__.py", line 100, in backward
        allow_unreachable=True)  # allow_unreachable flag
RuntimeError: Unable to find a valid cuDNN algorithm to run convolution (try_all at ..\aten\src\ATen\native\cudnn\Conv.cpp:693)
(no backtrace available)

win10 2080ti cuda10.2 cudnn7604 python 3.7 torch1.5 torchvision0.6

Trace screenshot:
what happend

Environment screenshot:
the enviroment

zst
  • 21
  • 4
  • can you come up with a simpler test case? If your cudnn is broke, even a simple 1 conv layer net with random inputs will break. Can you narrow the problem to a tiny reproducible example? – Shai May 13 '20 at 09:41

0 Answers0