[Non-Record] QAT Dead-Code Analysis + 7 Novel Technique Sweep (1xH100)

[wfproc]

Summary

Non-record research contribution on 1xH100:

• QAT dead-code bug: confirmed torch.compile constant-folds Late QAT in Record: 11L Partial RoPE + LN Scale + EMA + XSA4 (val_bpb: 1.1248) #315-derived code (present in SOTA Record: LeakyReLU² + Legal Score-First TTT + Parallel Muon — val_bpb 1.1194 (3-seed mean) #549). Tested three fix approaches, found a working tensor-scale STE method. Fixing it actually made the int6 gap worse, suggesting WD+EMA may already handle what QAT was supposed to do.
• 7 untried techniques swept: Muon-VS, Deep Delta Learning, Thinking Deeper recurrence, anti-layer removal, NS step reduction, spectral SVD compression, wider models. All negative.
• Key finding: at 83ms/step, each 1ms of overhead costs about 0.007 bpb. Only zero-overhead changes (like activation slope) survive at this step budget.

Also includes working implementations of prune-then-quantize and anti-layer diagnostic as env var toggles.

Full details in the README.

[MatoTeziTanka]

Community Review — [Non-Record] QAT Dead-Code Analysis + 7 Novel Technique Sweep (1xH100)

BPB: 0.007 (cache parse — may be delta/std, not val_bpb; check PR title) | Compliance: LOOKS CLEAN — score-first-per-chunk TTT (legal #1416/#1423 pattern)

What I found in the code (head SHA 8f3c4d531afe, file records/track_non_record_16mb/2026-03-28_QAT_DeadCode_Analysis_NovelTechniques_1xH100/train_gpt.py):

The TTT path at line 1202 implements the score-first-per-chunk pattern: each chunk is scored under torch.no_grad() / inference_mode() before the base_model.train() + SGD adaptation runs on that same chunk, with an is_last_chunk guard so the final chunk gets no adaptation pass. This is the structural shape the legal frontier uses (PRs #1416 erichroepke, #1423 aryanbhosale).

Per Issue #402 and Issue #677, TTT is legal when each token is scored before the adapter updates on it, and that's what the code does here — chunk ci is scored under weights adapted only on chunks 0..ci-1. No prequant_ttt_adapt_adamw(val_tokens, ...) multi-epoch fine-tune, no scored-region SLOT, no target-in-key n-gram cache.

CPU smoke test (CT2038 proteus-engine, 2026-04-11): import OK in 0.06s, dim=512, layers=11, vocab=1024, code=102945 B, SMOKE_TEST_PASS

Verdict: LOOKS CLEAN.

Recommendation to @cocohearts @valerio-oai @0hq @yuzhougu-oai @notapplica: MERGE pending standard checks (3-seed validation, 16MB artifact cap, 10-min wallclock on 8×H100 SXM). The compliance picture matches the legal reference frontier and no flags were raised by the classification pass.

Auto-classification caveat: this review was drafted by the AST-based classifier against a template derived from manually-reviewed cluster PRs (#1420, #1450, #1487, #1541, #1529, #1533, #1518). If I've misread a subtlety in your eval path — e.g., multi-epoch TTT that I mistook for single-pass, or a target-in-key lookup I missed in a helper function — please flag it and I'll re-run the audit manually.

---

Reviewed by @MatoTeziTanka — The Agora. CPU smoke test (CT2038 proteus-engine, 2026-04-11): import OK in 0.06s, dim=512, layers=11, vocab=1024, code=102945 B, SMOKE_TEST_PASS. Classification via deterministic AST-based classify_prs.py (pattern bank derived from ~65 manually-reviewed PRs earlier in the 2026-04-11 sweep). This review was auto-drafted from a template and spot-checked before posting — if the template misread your code, please call it out so I can iterate the classifier.