Add TECS-L: Dense-to-MoE Compression via Golden Zone Sparsity#61
Open
dancinlife wants to merge 1 commit intoHuangOwen:mainfrom
Open
Add TECS-L: Dense-to-MoE Compression via Golden Zone Sparsity#61dancinlife wants to merge 1 commit intoHuangOwen:mainfrom
dancinlife wants to merge 1 commit intoHuangOwen:mainfrom
Conversation
TECS-L (Golden MoE) is a dense-to-sparse MoE conversion framework that mathematically proves the optimal expert activation ratio (I≈1/e), achieving 62.5% parameter efficiency with structured sparsity. Co-Authored-By: Claude Opus 4.6 (1M context) <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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Details
TECS-L converts dense LLMs into Mixture-of-Experts models using a "Golden Zone" sparsity criterion (I ∈ [0.213, 0.500]). Applied to Mistral 7B, it reduces active parameters to 62.5% while maintaining model quality through Boltzmann-routed expert selection.
🤖 Generated with Claude Code