[OpenVINO] Add Support for GroupedMM in AWQ and Scale Estimation for OpenVINO Backend - #4176
Open
anzr299 wants to merge 12 commits into
Open
[OpenVINO] Add Support for GroupedMM in AWQ and Scale Estimation for OpenVINO Backend#4176anzr299 wants to merge 12 commits into
anzr299 wants to merge 12 commits into
Conversation
github-actions
Bot
added
documentation
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Changes
GroupedMatMul applies one weight matrix per group to a single shared 2D activation. Data-aware compression assumed a 3D weight always comes with a matching 3D activation, so AWQ, scale estimation and GPTQ failed.
AWQ
For weights, The experts axis is folded into the output channels, the search then minimizes the error over all experts at once, and the scale is broadcast back over the experts axis.
For activation, the same scale is applied over the hidden dim of the activation in non mergible case, For mergible scale, it is reshaped but still only applied over the hidden dim of the previous matmul weights.
Scale estimation
Broadcasts the pooled statistics over the group axis, so the scale stays per group.
Reason for changes
to support grouped_mm based MoE models in OpenVINO backend.
Related tickets
191895
Tests
test_scale_estimation and test_awq_scale_reference extended for this case.
test_node_utils extended for the weight channel axes and activation channel axis for GroupedMatMul.