[ROCm] Fix BlockwiseQuantLinear FP8 dtype handling for AMD GPUs#4136
Open
brucechanglongxu wants to merge 1 commit intopytorch:mainfrom
Open
[ROCm] Fix BlockwiseQuantLinear FP8 dtype handling for AMD GPUs#4136brucechanglongxu wants to merge 1 commit intopytorch:mainfrom
brucechanglongxu wants to merge 1 commit intopytorch:mainfrom
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/4136
Note: Links to docs will display an error until the docs builds have been completed. This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
bot
added
the
CLA Signed
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 join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
BlockwiseQuantLineardoesn't work on MI300 — the dtype whitelist rejects fnuz variants, the default is hardcoded toe4m3fn, and there's a bug whereself.dtypereads the class attribute (bfloat16) instead of storing the constructor arg, so the forward pass would always blow up infp8_blockwise_act_quant.Fixed by adding fnuz dtypes to the accepted set, defaulting to
e4m3_dtypefrom config (picks the right variant per hardware), and storing the dtype properly asself.fp8_dtype.Tested on MI300X — all four FP8 dtype variants construct fine, invalid dtypes get rejected, forward pass works.