all_results.json

{
    "epoch": 0.9221902017291066,
    "eval_logits/chosen": -0.0527331680059433,
    "eval_logits/rejected": 0.11135151982307434,
    "eval_logps/chosen": -138.20062255859375,
    "eval_logps/ref_chosen": -137.96424865722656,
    "eval_logps/ref_rejected": -174.7185821533203,
    "eval_logps/rejected": -174.97462463378906,
    "eval_loss": 0.6930533051490784,
    "eval_rewards/accuracies": 0.5249999761581421,
    "eval_rewards/chosen": -0.0023636952973902225,
    "eval_rewards/margins": 0.00019650773901958019,
    "eval_rewards/rejected": -0.0025602029636502266,
    "eval_runtime": 143.5558,
    "eval_samples": 1080,
    "eval_samples_per_second": 7.523,
    "eval_steps_per_second": 1.881,
    "total_flos": 0.0,
    "train_loss": 0.6934871196746826,
    "train_runtime": 628.3288,
    "train_samples": 1387,
    "train_samples_per_second": 2.207,
    "train_steps_per_second": 0.016
}