Qwen1.5-4B/train_results.json

{
    "epoch": 0.46,
    "train_loss": 0.584740468064944,
    "train_runtime": 28391.1424,
    "train_samples_per_second": 40.576,
    "train_steps_per_second": 0.159
}