Suggest setting the exact package version for torch in requirements.txt

Question

Suggest setting the exact package version for torch in requirements.txt

jmhb0 opened this issue a year ago · comments

I ran the install instructions on the README and tried running example_scripts/simple_example.py, but I got some error todo with the torch.nn.Upsample layer (sorry I forgot to save the error).

I was able to fix it by downgrading torch from >2.0 to 1.13.1. So maybe some issue was introduced when torch 2.0 launched.

I don't know if this is affecting others, but you might be able to avoid the issue by setting the exact torch version that you used in the requirements.txt.

Hirofumi Kobayashi · Answer 1 · Mon Oct 23 2023 14:44:48 GMT+0800 (China Standard Time)

Hi @jmhb0 thank you again for the bug report!
Yes, I also caught the error. The issue was that the size argument in Upsample must be int whereas I was passing numpy.int. The issue should be fixed in the PR. I tested it on pytorch 2.1.0. Let me know if it works for you too. Just FYI, bilinear upsampling may not be implemented in pytorch if you try to use 16bit precision or macOS MPS.

jmhb0 · Answer 2 · Tue Oct 31 2023 02:21:52 GMT+0800 (China Standard Time)

Thanks! That is a better fix :)