3fc9deac5b
- train_specialist_pt.py: pure PyTorch trainer replacing Fabricate binary. No SQLite/WAL — saves checkpoint.pt + history.json only. Fabricate was causing D-state IO blocking on the HDD due to continuous WAL writes. - eval_specialist_pt.py: PyTorch eval with confusion matrix, threshold analysis, guardrail assessment, and deployment verdict. - capacity_sweep.py: updated to call PyTorch scripts; TIERS now include heads field (4/6/8) since PyTorch supports multi-head unlike Fabricate. - prereq_op_binary_split.py: evaluate_combined rewritten in PyTorch. - Performance fix in train_specialist_pt.py: load training data onto GPU once and sample via torch.randint instead of DataLoader (eliminates Python/CPU overhead on tiny datasets). Fix applied, not yet timed. Next session: time the fix, then run sweep_all_gates.sh --pilot-only. See HANDOFF-2026-03-31.md for full context. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
11 lines
460 B
Plaintext
11 lines
460 B
Plaintext
=== run_grokking_until: whetstone_prereq_op ===
|
|
dataset=prereq_op_train.tsv lr=0.001 wd=0.01 max_steps=5000
|
|
Resuming from step 3000
|
|
Resuming: grokking already detected at step 1000, 3 blocks since grokking.
|
|
step= 4000 overall=62.5% T=62.5% F=62.5% loss=0.0000 min=0.0000 delta=+0.0% free=904GB <<< GROKKING (loss)
|
|
Stopping 4 blocks after grokking (step 4000).
|
|
|
|
=== DONE: whetstone_prereq_op ===
|
|
Final: step=4000 overall=62.5%
|
|
Grokking at step 1000
|