all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.0462207794189453,
    "eval_logits/rejected": -1.7758138179779053,
    "eval_logps/chosen": -90.40424346923828,
    "eval_logps/rejected": -73.17691802978516,
    "eval_loss": 0.5703843832015991,
    "eval_rewards/accuracies": 0.9472222328186035,
    "eval_rewards/chosen": 0.1581471860408783,
    "eval_rewards/margins": 0.2657936215400696,
    "eval_rewards/rejected": -0.10764642059803009,
    "eval_runtime": 118.208,
    "eval_samples": 2862,
    "eval_samples_per_second": 24.212,
    "eval_steps_per_second": 0.761,
    "train_loss": 0.6280729855576607,
    "train_runtime": 9689.6427,
    "train_samples": 140201,
    "train_samples_per_second": 14.469,
    "train_steps_per_second": 0.151
}