all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.2266557216644287,
    "eval_logits/rejected": -2.2218236923217773,
    "eval_logps/chosen": -34.07685852050781,
    "eval_logps/rejected": -37.5946159362793,
    "eval_loss": 0.6916552186012268,
    "eval_rewards/accuracies": 0.530315637588501,
    "eval_rewards/chosen": -0.0042304228991270065,
    "eval_rewards/margins": 0.0035695433616638184,
    "eval_rewards/rejected": -0.007799965795129538,
    "eval_runtime": 145.8282,
    "eval_samples": 343,
    "eval_samples_per_second": 2.352,
    "eval_steps_per_second": 0.295,
    "train_loss": 0.6806787961489195,
    "train_runtime": 3251.1509,
    "train_samples": 3079,
    "train_samples_per_second": 0.947,
    "train_steps_per_second": 0.118
}