the requirements.txt mandates Pytorch 2.0.1 to be used. However, this is not available from PyPI anymore. Furthermore, 2.0.1 seems to depend on CUDA 11.7/11.8, also not current. On the other hand, pull request #69 references Pytorch 2.2.1 for the base Docker image.
Please clarify. Will using 2.2.1 instead of 2.0.1 work?
@ndurner I believe the code is not strongly tied to the torch version; it might be an issue with xformer. You can try removing xformer and then try according to torch version 2.21? Looking forward to your feedback
Hi,
the requirements.txt mandates Pytorch 2.0.1 to be used. However, this is not available from PyPI anymore. Furthermore, 2.0.1 seems to depend on CUDA 11.7/11.8, also not current. On the other hand, pull request #69 references Pytorch 2.2.1 for the base Docker image.
Please clarify. Will using 2.2.1 instead of 2.0.1 work?
Thanks!