Record: SP8192 + Muon 0.97 + 3-Layer Recurrence + Parallel Residuals + TTT — val_bpb 1.0802 (3-seed mean)#1521
Open
aryanbhosale wants to merge 1 commit intoopenai:mainfrom
Conversation
…+ TTT — val_bpb 1.0802 (3-seed mean)
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Record: SP8192 + Muon 0.97 + 3-Layer Recurrence + Parallel Residuals + TTT
val_bpb = 1.0802 (3-seed mean, std 0.0007) | 8×H100 SXM
3-Seed Results
Merged SOTA (PR #1493): 1.0810. Delta: −0.0008 BPB.
Key Change
Single hyperparameter: Muon momentum 0.97 (from 0.99) on the merged #1 stack (PR #1493). Validated by PR #1514 (@dexhunter).
Full Stack
PR #1493 base + Muon 0.97: SP8192, MLP 4x, 3-layer depth recurrence (L3-5), parallel residuals (L7+), QK-Gain 5.25, MuonEq-R, WD=0.095, EMA=0.9965, warmdown=0.72, SDClip, GPTQ embeddings, score-first TTT (3 epochs), brotli.
Compliance (Track B)
Score-first TTT (PR #461). No SLOT, no pre-quant TTT, no n-gram. All conditions from Issue #1017 satisfied.
Reproduction
Credits
PR #1493 @bigbag, PR #1514 @dexhunter, PR #1394 @clarkkev, PR #1204 @msisovic