Qwen/Qwen1.5_1.8B_twitter/train_results.json

{
    "epoch": 3.0,
    "train_loss": 0.3625207788803998,
    "train_runtime": 885.2091,
    "train_samples": 8700,
    "train_samples_per_second": 29.485,
    "train_steps_per_second": 0.922
}