Qwen2.5-3B-grpo/train_results.json

{
    "total_flos": 0.0,
    "train_loss": 0.015861935917542785,
    "train_runtime": 47487.0739,
    "train_samples": 7498,
    "train_samples_per_second": 0.158,
    "train_steps_per_second": 0.005
}