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": 0.02,
      "learning_rate": 1.6666666666666664e-08,
      "logits/chosen": 0.22308249771595,
      "logits/rejected": 2.386685371398926,
      "logps/chosen": -637.75146484375,
      "logps/pi_response": -59.625553131103516,
      "logps/ref_response": -59.625553131103516,
      "logps/rejected": -297.7115783691406,
      "loss": 0.6931,
      "rewards/accuracies": 0.0,
      "rewards/chosen": 0.0,
      "rewards/margins": 0.0,
      "rewards/rejected": 0.0,
      "step": 1
    },
    {
      "epoch": 0.17,
      "learning_rate": 9.860114570402053e-08,
      "logits/chosen": 0.7615225315093994,
      "logits/rejected": 2.1085410118103027,
      "logps/chosen": -559.4776611328125,
      "logps/pi_response": -72.19044494628906,
      "logps/ref_response": -68.9166030883789,
      "logps/rejected": -288.6199645996094,
      "loss": 0.6639,
      "rewards/accuracies": 0.5972222089767456,
      "rewards/chosen": 0.042805079370737076,
      "rewards/margins": 0.06203806400299072,
      "rewards/rejected": -0.019232982769608498,
      "step": 10
    },
    {
      "epoch": 0.33,
      "learning_rate": 8.374915007591053e-08,
      "logits/chosen": 0.3314076066017151,
      "logits/rejected": 1.314693570137024,
      "logps/chosen": -558.7711791992188,
      "logps/pi_response": -875.009765625,
      "logps/ref_response": -72.45027923583984,
      "logps/rejected": -934.8829345703125,
      "loss": 0.3538,
      "rewards/accuracies": 0.793749988079071,
      "rewards/chosen": 0.5102975964546204,
      "rewards/margins": 7.6188764572143555,
      "rewards/rejected": -7.108579158782959,
      "step": 20
    },
    {
      "epoch": 0.5,
      "learning_rate": 5.738232820012406e-08,
      "logits/chosen": 0.21183089911937714,
      "logits/rejected": 0.8000686764717102,
      "logps/chosen": -564.4741821289062,
      "logps/pi_response": -3023.58251953125,
      "logps/ref_response": -92.48184204101562,
      "logps/rejected": -2812.103759765625,
      "loss": 0.2933,
      "rewards/accuracies": 0.7562500238418579,
      "rewards/chosen": 0.33371633291244507,
      "rewards/margins": 25.731952667236328,
      "rewards/rejected": -25.398237228393555,
      "step": 30
    },
    {
      "epoch": 0.67,
      "learning_rate": 2.8496739886173992e-08,
      "logits/chosen": 0.4574252963066101,
      "logits/rejected": 0.48074570298194885,
      "logps/chosen": -538.3971557617188,
      "logps/pi_response": -5107.06103515625,
      "logps/ref_response": -71.75224304199219,
      "logps/rejected": -4723.8134765625,
      "loss": 0.0916,
      "rewards/accuracies": 0.75,
      "rewards/chosen": 0.3258439898490906,
      "rewards/margins": 44.844539642333984,
      "rewards/rejected": -44.51869583129883,
      "step": 40
    },
    {
      "epoch": 0.84,
      "learning_rate": 6.947819411632222e-09,
      "logits/chosen": 0.4396973252296448,
      "logits/rejected": 0.41408687829971313,
      "logps/chosen": -516.14990234375,
      "logps/pi_response": -5759.0869140625,
      "logps/ref_response": -65.65478515625,
      "logps/rejected": -5587.03369140625,
      "loss": -0.0369,
      "rewards/accuracies": 0.8031250238418579,
      "rewards/chosen": 0.5476996898651123,
      "rewards/margins": 53.79410934448242,
      "rewards/rejected": -53.24641799926758,
      "step": 50
    },
    {
      "epoch": 0.99,
      "step": 59,
      "total_flos": 0.0,
      "train_loss": 0.22220247175733923,
      "train_runtime": 3779.3916,
      "train_samples_per_second": 4.044,
      "train_steps_per_second": 0.016
    }
  ],
  "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
}