d12 SFT v4 (num_iterations=700)

SFT on d12 pretrain checkpoint

Pretrain source: d12 H100 SSSL final (val_bpb=0.854, step 2205)

Plan:

1. Download pretrain checkpoint from MLflow artifacts (~2GB)
2. Train tokenizer (deterministic, ~80s)
3. Run SFT with default mixture (~1.07M rows, ~612 steps for full epoch)
4. Upload SFT checkpoint to MLflow artifacts for downstream RL
5. num_iterations=700 as cap (epoch-based stopping triggers at ~612)

Estimated: ~15 min on 1×H100, ~$3-5

Prior attempts: v1-v3 cancelled/failed (missing flag, YAML indent, inflated cost)

Status: dryrun pending