Out of memory error during evaluation but training works fine!

21

Awesome, thanks for taking the time to answer my questions.

23

Thanks for your comment, was helpful for me too! :slight_smile:

1 Like
24

It actually works. I’ve trained torchvision’s DensNet161 with 1 GPU. The model was trained well in training stage, but out of memory in eval. stage. Inference of model was performed without ‘torch.no_grad()’. Now there’s no out of memory. Thanks.

25

You’ve saved my homework and life. There’s full of tears of gratitude in my eyes. QAQ