DReP-SFT-Qwen3-4B-Thinking-2507/train_results.json

{
    "epoch": 4.0,
    "total_flos": 2.006101437057283e+19,
    "train_loss": 0.5864963867131946,
    "train_runtime": 153073.7159,
    "train_samples_per_second": 0.687,
    "train_steps_per_second": 0.005
}