RL-Forgetting-Experiments-3/qwen3-1.7b-base-code-sft-replay-hard-lam1-step102

q31p7bsftrphard_lam1

Inference-ready final coding-SFT model for q3_1p7b_sft_rp_hard_lam1 at optimizer step 102. Training uses qwen3_1p7b_s500_code_sft_data, order ordered, replay strategy hard, and replay lambda 1.0.

See delivery_manifest.json for immutable source lineage and file checksums.