Add linear_attn entries to Qwen3.5 base_model_tp_plan#47009
Open
ZAID646 wants to merge 1 commit into
Open
Conversation
Contributor
|
[For maintainers] Suggested jobs to run (before merge) run-slow: qwen3_5 |
Contributor
CI recapDashboard: View test results in Grafana |
Member
|
cc @Cyrilvallez for TP! |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Fixes #46846
Qwen3.5 is a hybrid model: ~75% of decoder layers use linear_attention (Gated DeltaNet) with their own projection matrices (
in_proj_qkv,in_proj_z,in_proj_b,in_proj_a,out_proj). These were missing frombase_model_tp_plan, causing:model.generate()(Conv1d channel mismatch afterin_proj_qkv)Fix: Add all
linear_attn.*projections with"colwise_gather_output"pattern, which shards the weight matrix (fixing OOM) and all-gathers activations before the depthwise Conv1d (fixing the shape mismatch).Updated both
modular_qwen3_5.py(source) andconfiguration_qwen3_5.py(auto-generated).