[moe] Add residual bottleneck variant for 10T gate experiments#4061
Open
claude[bot] wants to merge 1 commit intomainfrom
Open
[moe] Add residual bottleneck variant for 10T gate experiments#4061claude[bot] wants to merge 1 commit intomainfrom
claude[bot] wants to merge 1 commit intomainfrom
Conversation
…tion gates Create experiments/grug/moe_resid_bottleneck/ variant from MoE base adding per-layer learnable residual scaling (init 1.0) and per-head zero-init sigmoid attention gates. Both features are config-toggled. For the great 10T gate residual bottleneck experiments. Fixes #4035 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Contributor
|
🤖 Grug variant diff report
Artifact fallback: Download report bundle |
This was referenced Mar 30, 2026
Closed
Contributor
|
This pull request has been inactive for 23 days and is marked as stale. |
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Create experiments/grug/moe_resid_bottleneck/ variant from MoE base, adding per-layer learnable residual scaling (init 1.0, applied before each block) and per-head zero-init sigmoid attention gates (gate = 2 * sigmoid(W @ x[:12])). Both features are config-toggled via use_residual_lambdas and use_attention_gates flags. Existing grug variant contract tests auto-discover and validate the new variant.
Fixes #4035