search

ML-GSAI / DPT

Official PyTorch implementation for "Diffusion Models and Semi-Supervised Learners Benefit Mutually with Few Labels"
MIT License
78 stars 2 forks source link

the last step #5

Closed holydick99 closed 2 months ago

holydick99 commented 3 months ago

the last step it look same as the first step but when i input the command line it always runtime error. @K5_XZ38F}S_U9~I_BY %%X

yyyouy commented 3 months ago

thank you for your attention. try this: https://github.com/THUDM/ChatGLM-6B/pull/1173/files maybe this is because of the problem of version.

holydick99 commented 3 months ago

sorry about that im a green-hand of this area, so this solution means use these code to replace the original code?

yyyouy commented 3 months ago

HH, don't feel sorry, I know issues with the environment can be frustrating. What I mean is, try adding the code included to see if it helps. It might be due to torchrun. Try launching with torchrun and also add the code from the link at the beginning of your training script.

holydick99 commented 3 months ago

$Y0S}80(LIB441R2B5LL 6L so this specific is change this code?

yyyouy commented 3 months ago

yes, you can have a try.

holydick99 commented 3 months ago

HH, don't feel sorry, I know issues with the environment can be frustrating. What I mean is, try adding the code included to see if it helps. It might be due to torchrun. Try launching with torchrun and also add the code from the link at the beginning of your training script.

really thank u, the right way is replace that code to (torch.distributed.init_process_group), then it can run normally

yyyouy commented 3 months ago

HH, don't feel sorry, I know issues with the environment can be frustrating. What I mean is, try adding the code included to see if it helps. It might be due to torchrun. Try launching with torchrun and also add the code from the link at the beginning of your training script.

really thank u, the right way is replace that code to (torch.distributed.init_process_group), then it can run normally

very glad to hear this method can help you.