trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.9922480620155039,
  "eval_steps": 100,
  "global_step": 64,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.02,
      "grad_norm": 5.753161071734387,
      "learning_rate": 7.142857142857142e-08,
      "logits/chosen": -1.146815299987793,
      "logits/rejected": -1.0666704177856445,
      "logps/chosen": -606.579345703125,
      "logps/rejected": -505.8602294921875,
      "loss": 0.6931,
      "rewards/accuracies": 0.0,
      "rewards/chosen": 0.0,
      "rewards/margins": 0.0,
      "rewards/rejected": 0.0,
      "step": 1
    },
    {
      "epoch": 0.16,
      "grad_norm": 5.7493431535662305,
      "learning_rate": 4.965903258506806e-07,
      "logits/chosen": -1.1877351999282837,
      "logits/rejected": -1.1102757453918457,
      "logps/chosen": -546.2195434570312,
      "logps/rejected": -661.2600708007812,
      "loss": 0.693,
      "rewards/accuracies": 0.5277777910232544,
      "rewards/chosen": 0.0011070389300584793,
      "rewards/margins": 0.001627296325750649,
      "rewards/rejected": -0.0005202573956921697,
      "step": 10
    },
    {
      "epoch": 0.31,
      "grad_norm": 5.1113110991031645,
      "learning_rate": 4.38526652444224e-07,
      "logits/chosen": -1.182856559753418,
      "logits/rejected": -1.096764326095581,
      "logps/chosen": -496.00006103515625,
      "logps/rejected": -631.6556396484375,
      "loss": 0.6878,
      "rewards/accuracies": 0.8125,
      "rewards/chosen": 0.011512617580592632,
      "rewards/margins": 0.011010391637682915,
      "rewards/rejected": 0.0005022261175327003,
      "step": 20
    },
    {
      "epoch": 0.47,
      "grad_norm": 5.707953313171029,
      "learning_rate": 3.2462870275042367e-07,
      "logits/chosen": -1.2160546779632568,
      "logits/rejected": -1.1250536441802979,
      "logps/chosen": -462.3262634277344,
      "logps/rejected": -688.6295166015625,
      "loss": 0.6789,
      "rewards/accuracies": 0.8374999761581421,
      "rewards/chosen": 0.03221649304032326,
      "rewards/margins": 0.029845599085092545,
      "rewards/rejected": 0.0023708937224000692,
      "step": 30
    },
    {
      "epoch": 0.62,
      "grad_norm": 5.28907700413882,
      "learning_rate": 1.886286282148002e-07,
      "logits/chosen": -1.2214853763580322,
      "logits/rejected": -1.1557971239089966,
      "logps/chosen": -473.38720703125,
      "logps/rejected": -619.6802978515625,
      "loss": 0.671,
      "rewards/accuracies": 0.8812500238418579,
      "rewards/chosen": 0.049382228404283524,
      "rewards/margins": 0.04463439807295799,
      "rewards/rejected": 0.004747826140373945,
      "step": 40
    },
    {
      "epoch": 0.78,
      "grad_norm": 5.6548124405750775,
      "learning_rate": 7.080437170788722e-08,
      "logits/chosen": -1.2320277690887451,
      "logits/rejected": -1.094261884689331,
      "logps/chosen": -481.0409240722656,
      "logps/rejected": -643.069091796875,
      "loss": 0.6664,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 0.05804261565208435,
      "rewards/margins": 0.053116798400878906,
      "rewards/rejected": 0.0049258084036409855,
      "step": 50
    },
    {
      "epoch": 0.93,
      "grad_norm": 5.6228626485550155,
      "learning_rate": 6.0509043431410945e-09,
      "logits/chosen": -1.2452279329299927,
      "logits/rejected": -1.109863042831421,
      "logps/chosen": -473.0909729003906,
      "logps/rejected": -635.895263671875,
      "loss": 0.6614,
      "rewards/accuracies": 0.8999999761581421,
      "rewards/chosen": 0.05887182429432869,
      "rewards/margins": 0.060193151235580444,
      "rewards/rejected": -0.0013213239144533873,
      "step": 60
    },
    {
      "epoch": 0.99,
      "step": 64,
      "total_flos": 0.0,
      "train_loss": 0.6759306937456131,
      "train_runtime": 862.557,
      "train_samples_per_second": 4.756,
      "train_steps_per_second": 0.074
    }
  ],
  "logging_steps": 10,
  "max_steps": 64,
  "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
}