Wang, Yi
Wang, Yi
@yao-matrix @sgugger please notice the issue and help review
@sgugger I agree with your point, so I upload another PR (https://github.com/huggingface/transformers/pull/19151) to only update the doc.
we will release a new oneCCL 1.12.1 to work with torch 1.12.1
@sgugger @yao-matrix same with the PR with transformer
@sgugger we will release a new oneCCL 1.12.1 to work with torch 1.12.1
thanks very much @Narsil and @OlivierDehaene for your comment. adding @yao-matrix
@Narsil @OlivierDehaene I add xpu smi in env runtime, do you think is it proper to add this? this is to dump intel XPU version.
@Narsil could you review the PR?done by my side.
@Narsil @mfuntowicz CI node is ready, could we move on?
@pacman100 could you help review? found the issue in FSDP(cpu offload)+PEFT