hpcaitech/ColossalAI

[BUG]: Torch compile causes multi-process to hang with python 3.9

Opened this issue · 0 comments

Is there an existing issue for this bug?

  • I have searched the existing issues

🐛 Describe the bug

Posting this for documentation purposes.
torch.compile has been reported to cause hang on exit when used with the "spawn" method in multiprocess.
pytorch/pytorch#115366
pytorch/pytorch#126311
Upgrading to Python 3.10 may solve the problem (at least in my case). Currently, we can only push the Pytorch community to fix this🫠

Environment

PyTorch 2.1.1 + Python 3.9