[WS2][GEMM][3th merge] feat: add wrapper to ablation matrix - #322
Merged
Conversation
Keep the FP32 running sum for tensor-core speed, and only quantize once at store time.
Keep FP32 leaves for tensor-core speed and BF16-add internal nodes. TP=2 is one commutative add; TP=8 left-fold is a different tree and is expected to diverge.
Replace NCCL with fixed-tree collectives and AllGather CP tokens before det_gemm_db so weight grads match CP=1. Fold FFN coverage into tests/test_qwen_ffn.py.
Default True keeps det_gemm with no split-K. False uses torch.matmul so callers can take the cuBLASLt/CUTLASS path.
frank-2077
requested review from
Flink-ddd,
KJLdefeated,
bitborne and
inaniloquentee
as code owners
August 19, 2026 14:20
|
Important Review skippedAuto reviews are disabled on base/target branches other than the default branch. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Plus Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
maxiaosong1124
self-requested a review
August 24, 2026 15:58
maxiaosong1124
approved these changes
Aug 24, 2026
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
3 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Expose a Qwen3 dense FFN wrapper so the ablation matrix can switch GEMM backends.
disable_split_k=True: deterministicdet_gemm(no split-K), TP/CP/SP with fixed-tree collectives.disable_split_k=False:torch.matmul(cuBLASLt / CUTLASS, split-K allowed).