SwiGLU MLP: parameter-neutral gated activation over LeakyReLU^2#676
Open
they-call-me-god wants to merge 1 commit intoopenai:mainfrom
Open
SwiGLU MLP: parameter-neutral gated activation over LeakyReLU^2#676they-call-me-god wants to merge 1 commit intoopenai:mainfrom
they-call-me-god wants to merge 1 commit intoopenai:mainfrom
Conversation
Replace LeakyReLU(0.5)^2 with SwiGLU (silu gate * up projection). Same parameter count: 3*512*1024 = 2*512*1536 = 1,572,864 per layer. All other SOTA settings preserved (TTT, Parallel Muon, int6+lzma, etc.) Co-Authored-By: Claude Sonnet 4.6 <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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Replace
LeakyReLU(0.5)^2with SwiGLU gating — the same multiplicative activation used in LLaMA, Mistral, Gemma, and PaLM.Built on the PR #549 SOTA stack (LeakyReLU² + Legal TTT + Parallel Muon). Single change, zero parameter increase.
The Change
Parameter Neutrality
Proof:
2 × 512 × 1536 = 3 × 512 × 1024 = 1,572,864per layer.Status
Training logs pending (RunPod 8×H100). Will update with 3-seed results and final val_bpb.
New env vars
USE_SWIGLU=1(default on)SWIGLU_HALF_DIM=1024