Fix: Pin OneTrainer ROCm bitsandbytes wheel to 0.49.1 - #1723
Merged
Conversation
|
CLA Assistant Lite bot CLA Assistant bot All Contributors have signed the CLA. |
Contributor
Author
|
I have read the CLA Document and I hereby sign the CLA |
0xDELUXA
force-pushed
the
onetrainer-bnb-0.50.2
branch
from
August 27, 2026 21:41
5ddbf1b to
258270c
Compare
Contributor
|
LGTM |
mohnjiles
approved these changes
Aug 29, 2026
ionite34
approved these changes
Aug 29, 2026
Member
|
Thank you! |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes #1708.
Points
OneTrainerWindowsRocmProfileat the0.49.1wheel built for HIP 7.13-7.16, replacing the0.50.0.dev0.The profile currently installs a
0.50wheel, but OneTrainer pinsbitsandbytes==0.49.1and passesblock_wise=andpercentile_clipping=to every bnb optimizer.0.50removed both arguments, so any training run usingAdamW8bitand friends dies at startup with aTypeError. Upstream OneTrainer is not moving to0.50for now (Nerogar/OneTrainer#1708 was closed), so matching the pin is the fix.The wheel covers 27 gfx targets. Verified on Windows with HIP 7.16: 4-bit linear, 8-bit linear and an 8-bit optimizer step all run, and the version now matches OneTrainer's pin exactly so pip has nothing to resolve.
cc @NeuralFault