Non-record: Mac mini M4 16GB, no H100s, still golfing (val_bpb=1.5672)

[frido22]

Summary

• Non-record Apple Silicon / MLX submission
• Hardware: Mac mini M4 16GB
• Verified best published run from frido22/low_vram_institute
• Run ID: 2026_03_24_run_0033
• Run title: Late Projection EMA Finalizer
• Final exact post-quant score: val_bpb = 1.56720003
• Final exact post-quant loss: val_loss = 3.53539760
• Total artifact size: 15,962,372 bytes
• Int8+zlib model size: 15,888,695 bytes

Notes

This is not an official 8xH100 record submission. It is intentionally submitted under records/track_non_record_16mb as a hardware-specific Apple Silicon non-record entry.

The submission folder was created from the published run package only:

• output/runs/2026_03_24_run_0033

Verification was done against the published repo state, including:

• state/ledger.jsonl
• output/reports/history.csv
• output/runs/2026_03_24_run_0033/submission.json
• output/runs/2026_03_24_run_0033/artifact_size.json
• output/runs/2026_03_24_run_0033/run.log

[MatoTeziTanka]

Community Review — Non-record: Mac mini M4 16GB, no H100s, still golfing (val_bpb=1.5672)

Compliance: NEEDS AUTHOR ACTION — train_gpt.py fails to import on CT2038 (Python 3.10 / torch 2.10.0+cpu)

What I found: The CPU smoke test on CT2038 (proteus-engine, 128 GB RAM, Triton 3.6.0, flash_attn stub, cutlass_evt_fusion stub) failed at the import step with:

ModuleNotFoundError: No module named 'mlx'

A few of the common patterns I've seen for this class of error in the 2026-04-11 sweep:

• PEP 701 f-string nesting — e.g. log(f" {cat}: {", ".join(...)}") is valid Python 3.12+ but invalid Python 3.10 because the inner ", " re-enters the outer double-quote context. One-character fix: ', ' instead of ", ". See PR Record: SP8192 + Improved Parallel Residuals + Muon 0.97 + LR 0.03 + Legal TTT — val_bpb 1.07785 (3-seed mean) #1541 / Record: SP8192 + Triple Recurrence + Banking + Fused MLP + Muon 0.97 — val_bpb 1.0778 (3-seed mean) #1523 for reference.
• Missing flash_attn variants — e.g. from flash_attn_interface import flash_attn_varlen_func when the wrapper script only stubs flash_attn_func. Not a PR defect on H100s, but the eval image / CPU preflight path needs a guarded import.
• Local compiled extension — e.g. import cutlass_evt_fusion from a records/*/cutlass_evt_fusion/ subfolder that isn't on the import path at smoke time. Usually an import-order issue inside the script.
• Actual syntax error — typo, missing bracket, etc.

Recommendation: Could you run python3 -c "import py_compile; py_compile.compile('train_gpt.py')" on your records-folder train_gpt.py under Python 3.10 specifically? The eval image is Python 3.10 per Issue #17 / the README, so any parse error on 3.10 blocks the submission at import time before any of the scored-eval logic runs.

Once the parse/import issue is fixed, I'll re-run the compliance audit through the normal pipeline. No other flags identified yet because the audit halts at the import step.

---

Reviewed by @MatoTeziTanka — The Agora. CPU smoke test (CT2038 proteus-engine, 2026-04-11): IMPORT_FAIL — ModuleNotFoundError: No module named 'mlx'. Classification via classify_prs.py AST-based classifier; full compliance audit deferred until the import issue is resolved. Auto-drafted from a template and spot-checked before posting.

[frido22]

Superseded by #1762, which carries the updated Mac mini M4 non-record submission, the improved 1.51996743 result, and the records-folder import hardening for the earlier missing-mlx CPU smoke issue.