[bnb] Move documentation #18671
[bnb] Move documentation #18671
Conversation
- move bnb documentation to `infer_gpu_many`
|
The documentation is not available anymore as the PR was closed or merged. |
|
yeah, I wasn't sure, probably you're right and then like perf_train_gpu_many.mdx says on top to first read perf_train_gpu_one.mdx - add the same to perf_infer_gpu_many.mdx? |
- added text on infer_gpu_one - added a small note on infer_gpu_many - added customized multi gpu example on infer_gpu_many
|
Yep makes sense! I propose a small refactoring at 2018285 ! |
|
The proposed change would be difficult to maintain and 2 copies will get out of sync. Only one copy please - if you prefer the one gpu doc that's where it should be. the other one linking to it. |
Co-authored-by: Stas Bekman <stas00@users.noreply.github.com>
Co-authored-by: Stas Bekman <stas00@users.noreply.github.com>
|
Proposed a change in 5f8a3ae ! Let me know if this works for you |
stas00
left a comment
There was a problem hiding this comment.
with a few tweaks I proposed this will now look good to merge.
thank you, @younesbelkada for working on it!
|
Thanks a lot @stas00 for iterating on the changes 💪 |
Co-authored-by: Stas Bekman <stas00@users.noreply.github.com>
What does this PR do?
perf_infer_gpu_many.mdxas it was previously set toperf_train_gpu_one.mdxwhich is not relevant in the case ofbitsandbytesintegration since it supports inference only.cc @stas00
I do have a question though, what about
perf_infer_gpu_one.mdx? I think that thebnbdocumentation could fit well in this file as well since it supports single GPU inference too.