Update SFT DoRA adapter (lora_best.pt + PEFT format)

SFT retrain 2026-07-07: repaired edit-grounding corpus (0 ungrounded, 58 recovery demos) + lean native prompt (1983 tok prefix); val_loss 0.9159

FIX: PEFT key format — add base_model.model. prefix (adapter was silently unapplied by PeftModel.from_pretrained)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format) — 2026-05-18, val_loss=0.62025

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Sync SFT adapter from RL policy_v1

Sync SFT adapter from RL policy_v1

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Update SFT DoRA adapter (lora_best.pt + PEFT format)

Upload SFT DoRA adapter (final, val_loss=0.2681)

Upload SFT DoRA adapter (best val_loss=0.2671)

Update SFT DoRA adapter final checkpoint

Update SFT DoRA adapter (best val_loss=0.2690, 3 epochs on 628 samples)

Upload folder using huggingface_hub