[perf] fuse AllReduce + RMSNorm + FP8 quant for ds/kimi#1388
Open
gbyu-amd wants to merge 5 commits into
Open
Conversation
…rm (#1226) The combined HIP kernel emits the quantized (fp8, scale) activation that the downstream qkv GEMM consumes directly, removing a standalone per-token/per-group quant kernel from the hot path. - layernorm.py: RMSNorm.forward gains a fused AR+RMS+quant branch dispatching to `..._rmsnorm_quant_per_group` (per_1x128) or `..._rmsnorm_quant` (per_Token), returning ((fp8, scale), residual). - deepseek_v2.py: enable `fused_quant` on input_layernorm when AR fusion is on and fused_qkv_a_proj is per_1x128/per_Token FP8. Mutually exclusive with the existing non-AR `fuse_input_norm_quant` path; the attention forward already unpacks the (fp8, scale) tuple. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
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
2 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Motivation
The combined HIP kernel emits the quantized (fp8, scale) activation that the downstream qkv GEMM consumes directly, removing a standalone per-token/per-group quant kernel from the hot path.
..._rmsnorm_quant_per_group(per_1x128) or..._rmsnorm_quant(per_Token), returning ((fp8, scale), residual).fused_quanton input_layernorm when AR fusion is on and fused_qkv_a_proj is per_1x128/per_Token FP8. Mutually exclusive with the existing non-ARfuse_input_norm_quantpath; the attention forward already unpacks the (fp8, scale) tuple.Need the fix from aiter: ROCm/aiter#3977
Technical Details
Test Plan
Test Result
Submission Checklist