web_rl_3b/train_results.json

{
    "epoch": 1.0,
    "total_flos": 57141526265856.0,
    "train_loss": 0.2802479102018294,
    "train_runtime": 34553.4877,
    "train_samples_per_second": 0.246,
    "train_steps_per_second": 0.031
}