I trained LoRA adapters for Qwen3-4B using Unsloth on mixed instruction data.I’m looking for feedback on instruction following and stability.
· Sign up or log in to comment