trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.9874476987447699,
  "eval_steps": 500,
  "global_step": 59,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.016736401673640166,
      "grad_norm": 117.7337967450897,
      "learning_rate": 8.333333333333333e-08,
      "logits/chosen": 0.5924187898635864,
      "logits/rejected": 2.4197378158569336,
      "logps/chosen": -481.50433349609375,
      "logps/pi_response": -57.85610580444336,
      "logps/ref_response": -57.85610580444336,
      "logps/rejected": -407.123291015625,
      "loss": 0.6931,
      "rewards/accuracies": 0.0,
      "rewards/chosen": 0.0,
      "rewards/margins": 0.0,
      "rewards/rejected": 0.0,
      "step": 1
    },
    {
      "epoch": 0.16736401673640167,
      "grad_norm": 114.88052445634541,
      "learning_rate": 4.930057285201027e-07,
      "logits/chosen": 1.0167477130889893,
      "logits/rejected": 1.715598464012146,
      "logps/chosen": -461.42584228515625,
      "logps/pi_response": -612.001708984375,
      "logps/ref_response": -43.13614273071289,
      "logps/rejected": -751.348388671875,
      "loss": 0.5847,
      "rewards/accuracies": 0.6041666865348816,
      "rewards/chosen": 0.18106825649738312,
      "rewards/margins": 5.178502082824707,
      "rewards/rejected": -4.997433185577393,
      "step": 10
    },
    {
      "epoch": 0.33472803347280333,
      "grad_norm": 74.42064867560319,
      "learning_rate": 4.187457503795526e-07,
      "logits/chosen": 0.9024032354354858,
      "logits/rejected": 1.7998549938201904,
      "logps/chosen": -459.34100341796875,
      "logps/pi_response": -839.3756713867188,
      "logps/ref_response": -48.8006591796875,
      "logps/rejected": -996.8543090820312,
      "loss": 0.3541,
      "rewards/accuracies": 0.793749988079071,
      "rewards/chosen": 0.46666306257247925,
      "rewards/margins": 7.785676002502441,
      "rewards/rejected": -7.319011688232422,
      "step": 20
    },
    {
      "epoch": 0.502092050209205,
      "grad_norm": 130.4626377347778,
      "learning_rate": 2.8691164100062034e-07,
      "logits/chosen": 2.6123948097229004,
      "logits/rejected": 3.322319507598877,
      "logps/chosen": -594.0442504882812,
      "logps/pi_response": -878.4356689453125,
      "logps/ref_response": -42.91078567504883,
      "logps/rejected": -1136.06591796875,
      "loss": 0.2937,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -1.1390485763549805,
      "rewards/margins": 7.548802852630615,
      "rewards/rejected": -8.68785285949707,
      "step": 30
    },
    {
      "epoch": 0.6694560669456067,
      "grad_norm": 53.877472153382854,
      "learning_rate": 1.4248369943086995e-07,
      "logits/chosen": 2.7364916801452637,
      "logits/rejected": 3.1835875511169434,
      "logps/chosen": -542.8189697265625,
      "logps/pi_response": -965.1339721679688,
      "logps/ref_response": -44.70442581176758,
      "logps/rejected": -1189.79150390625,
      "loss": 0.2283,
      "rewards/accuracies": 0.9312499761581421,
      "rewards/chosen": -0.6533902883529663,
      "rewards/margins": 8.359137535095215,
      "rewards/rejected": -9.012528419494629,
      "step": 40
    },
    {
      "epoch": 0.8368200836820083,
      "grad_norm": 56.17404888873974,
      "learning_rate": 3.473909705816111e-08,
      "logits/chosen": 2.6901650428771973,
      "logits/rejected": 3.5459206104278564,
      "logps/chosen": -530.8473510742188,
      "logps/pi_response": -472.94268798828125,
      "logps/ref_response": -48.817832946777344,
      "logps/rejected": -734.7946166992188,
      "loss": 0.2278,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": -0.4413676857948303,
      "rewards/margins": 4.092815399169922,
      "rewards/rejected": -4.534182548522949,
      "step": 50
    },
    {
      "epoch": 0.9874476987447699,
      "step": 59,
      "total_flos": 0.0,
      "train_loss": 0.3167330996464875,
      "train_runtime": 2832.9137,
      "train_samples_per_second": 5.395,
      "train_steps_per_second": 0.021
    }
  ],
  "logging_steps": 10,
  "max_steps": 59,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 100,
  "total_flos": 0.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}