qwen3-4b-dabstep-reasoning-108-fixed-reasoning-sharegpt-sft/train_results.json

{
    "epoch": 5.0,
    "total_flos": 9984310984704.0,
    "train_loss": 0.5776977070740291,
    "train_runtime": 1052.4329,
    "train_samples_per_second": 0.48,
    "train_steps_per_second": 0.033
}