fix(0306): MoE prefill reductions for subfunction export#1028
Open
vbaddi wants to merge 2 commits into
Open
Conversation
vbaddi
force-pushed
the
fix/subfunction-reducesum-einsum
branch
from
829f6e9 to
d44816b
Compare
Replace ReduceSum-prone MoE prefill aggregation paths with equivalent einsum reductions for GLM4-MoE, Qwen3-MoE, and GPT-OSS. Add tiny-model ONNX subfunction quickcheck coverage to verify the exported decoder subfunctions include Einsum and expected MoE custom ops. Signed-off-by: vbaddi <vbaddi@qti.qualcomm.com>
Revert Qwen3-MoE top-k probability normalization back to the original .sum(-1, keepdim=True) path, keeping only the prefill expert-output reduction as einsum. Targeted Qwen3-MoE subfunction and parity tests pass. Signed-off-by: vbaddi <vbaddi@qti.qualcomm.com>
vbaddi
force-pushed
the
fix/subfunction-reducesum-einsum
branch
from
d44816b to
d75e27d
Compare
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Replaces selected MoE prefill and expert aggregation reductions with equivalent einsum forms for GPT-OSS, GLM-MOE and Qwen3-MoE. Adds tiny-model ONNX subfunction quickchecks that verify decoder subfunctions contain einsum.
cc: @quic-rishinr @mohiso22