GRPO trained LoRA model based on unsloth/Qwen3-4B (Trained with Unsloth) 2f3dd40 verified thejaminator commited on Jun 2, 2025