all_results.json

{
    "epoch": 0.9880609304240429,
    "eval_logits/chosen": 9.247183799743652,
    "eval_logits/rejected": 10.069817543029785,
    "eval_logps/chosen": -0.4010205566883087,
    "eval_logps/rejected": -0.41174402832984924,
    "eval_loss": 1.3837605714797974,
    "eval_rewards/accuracies": 0.5231481194496155,
    "eval_rewards/chosen": -0.8020411133766174,
    "eval_rewards/margins": 0.02144695073366165,
    "eval_rewards/rejected": -0.8234880566596985,
    "eval_runtime": 145.5386,
    "eval_samples": 1080,
    "eval_samples_per_second": 7.421,
    "eval_steps_per_second": 1.855,
    "total_flos": 0.0,
    "train_loss": 1.401187040011088,
    "train_runtime": 7987.3887,
    "train_samples": 9715,
    "train_samples_per_second": 1.216,
    "train_steps_per_second": 0.009
}