Tralalabs

/

Nora-4B-LoRA

like 1

I trained LoRA adapters for Qwen3-4B using Unsloth on mixed instruction data. I’m looking for feedback on instruction following and stability.