Skip to content

Record: SP8192 + Improved Parallel Residuals + Muon 0.97 + TTT 5ep + N-gram Tilt + Hessian SDClip — val_bpb 1.07730 #1557

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
ndokutovich wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Record: SP8192 + Improved Parallel Residuals + Muon 0.97 + TTT 5ep + N-gram Tilt + Hessian SDClip — val_bpb 1.07730 #1557

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
38 changes: 38 additions & 0 deletions ...2026-04-12_SP8192_ImprovedParResid_Muon097_TTT5ep_NgramTilt_HessianSDClip/submission.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,38 @@
{
"author": "ndokutovich",
"github_id": "ndokutovich",
"name": "SP8192 + Improved Parallel Residuals + Muon 0.97 + LR 0.03 + Score-First TTT 5ep + Causal N-gram Tilt + Hessian SDClip",
"date": "2026-04-12",
"track": "10min_16mb",
"val_bpb": 1.07730,
"val_bpb_std": 0.00040,
"seeds": [42, 314, 999],
"seed_results": {
"42": {"val_bpb": 1.07684, "artifact_bytes": 15965495},
"314": {"val_bpb": 1.07748, "artifact_bytes": 15965495},
"999": {"val_bpb": 1.07757, "artifact_bytes": 15965495}
},
"hardware": "8xH100 80GB SXM",
"pytorch_version": "2.4.0",
"technique_summary": "SP8192 + Improved Parallel Residuals (cross-lane L7+) + 3-Layer Depth Recurrence (L3-5) + Muon 0.97 + Matrix LR 0.03 + EMA 0.997 + QK-Gain 5.0 + Score-First TTT (SGD 5ep lr=0.005) + Causal N-gram Tilt (beta=2.0 agree=0.1) + Hessian-Aware SDClip (lambda=0.175) + GPTQ int6/int8 + Brotli",
"compliance": {
"train_under_600s": true,
"artifact_under_16mb": true,
"eval_under_600s": true,
"no_slot": true,
"no_pre_quant_ttt": true,
"no_etlb": true,
"no_ngram_cache": true,
"no_hash_embed": true,
"score_first_ttt": true,
"causal_ngram_tilt": true,
"three_seeds": true
},
"attribution": {
"base": "PR #1529 msisovic (1.0753 BPB) — improved parallel residuals architecture",
"ttt": "Score-first TTT from PR #461 framework",
"ngram_tilt": "Causal n-gram tilt from PR #1514 AnirudhRahul",
"hessian_sdclip": "Hessian-Aware SDClip from PR #1412 Robby955",
"optimizer": "Muon 0.97 + LR 0.03 from PR #1493 bigbag tuning"
}
}
Loading