Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA) — 2026-05-18, val_loss=0.62025

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Sync SFT merged model from RL training (policy_v1)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update merged SFT model + adapter (Qwen3-Coder-30B-A3B + DoRA)

Update SFT DoRA adapter: best val_loss=0.2690, 3 epochs, targets=[q,k,v,o_proj], r=16, alpha=32 (raw pt)

Update SFT DoRA adapter: best val_loss=0.2690, 3 epochs, targets=[q,k,v,o_proj], r=16, alpha=32 (config)

Update SFT DoRA adapter: best val_loss=0.2690, 3 epochs, targets=[q,k,v,o_proj], r=16, alpha=32 (safetensors)

Upload folder using huggingface_hub

Delete lora_final.pt with huggingface_hub

Upload lora_best.pt with huggingface_hub

Add original lora_best.pt checkpoint

Upload DoRA adapter (r=16, q_proj+v_proj) trained on debugger SFT data

Upload lora_final.pt with huggingface_hub