Qwen2.5-7B-ODA-Mixture-100k/train_results.json

{
    "epoch": 3.0,
    "total_flos": 6.364819972765516e+18,
    "train_loss": 0.2649726264235014,
    "train_runtime": 208629.7363,
    "train_samples_per_second": 0.569,
    "train_steps_per_second": 0.018
}