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": 3.1506166106625195,
      "learning_rate": 1.6666666666666664e-08,
      "logits/chosen": -0.16345731914043427,
      "logits/rejected": -0.16345731914043427,
      "logps/chosen": -1051.055419921875,
      "logps/pi_response": -1051.055419921875,
      "logps/ref_response": -1051.055419921875,
      "logps/rejected": -1051.055419921875,
      "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": 34.149713113458255,
      "learning_rate": 9.860114570402053e-08,
      "logits/chosen": -0.17107006907463074,
      "logits/rejected": -0.17107006907463074,
      "logps/chosen": -1237.6085205078125,
      "logps/pi_response": -1237.5712890625,
      "logps/ref_response": -1047.64404296875,
      "logps/rejected": -1237.6085205078125,
      "loss": 0.6931,
      "rewards/accuracies": 0.0,
      "rewards/chosen": -1.8992481231689453,
      "rewards/margins": 0.0,
      "rewards/rejected": -1.8992481231689453,
      "step": 10
    },
    {
      "epoch": 0.33472803347280333,
      "grad_norm": 89.9356708413712,
      "learning_rate": 8.374915007591053e-08,
      "logits/chosen": -0.8497294187545776,
      "logits/rejected": -0.8497294187545776,
      "logps/chosen": -6867.37353515625,
      "logps/pi_response": -6867.2529296875,
      "logps/ref_response": -1047.505615234375,
      "logps/rejected": -6867.37353515625,
      "loss": 0.6931,
      "rewards/accuracies": 0.0,
      "rewards/chosen": -58.19902420043945,
      "rewards/margins": 0.0,
      "rewards/rejected": -58.19902420043945,
      "step": 20
    },
    {
      "epoch": 0.502092050209205,
      "grad_norm": 93.52643691744206,
      "learning_rate": 5.738232820012406e-08,
      "logits/chosen": -1.5894163846969604,
      "logits/rejected": -1.5894163846969604,
      "logps/chosen": -19478.552734375,
      "logps/pi_response": -19478.759765625,
      "logps/ref_response": -1048.0611572265625,
      "logps/rejected": -19478.552734375,
      "loss": 0.6931,
      "rewards/accuracies": 0.0,
      "rewards/chosen": -184.30494689941406,
      "rewards/margins": 0.0,
      "rewards/rejected": -184.30494689941406,
      "step": 30
    },
    {
      "epoch": 0.6694560669456067,
      "grad_norm": 68.26794136278654,
      "learning_rate": 2.8496739886173992e-08,
      "logits/chosen": -1.8664979934692383,
      "logits/rejected": -1.8664979934692383,
      "logps/chosen": -30769.759765625,
      "logps/pi_response": -30769.974609375,
      "logps/ref_response": -1046.073486328125,
      "logps/rejected": -30769.759765625,
      "loss": 0.6931,
      "rewards/accuracies": 0.0,
      "rewards/chosen": -297.2370300292969,
      "rewards/margins": 0.0,
      "rewards/rejected": -297.2370300292969,
      "step": 40
    },
    {
      "epoch": 0.8368200836820083,
      "grad_norm": 67.0066669202387,
      "learning_rate": 6.947819411632222e-09,
      "logits/chosen": -2.048996686935425,
      "logits/rejected": -2.048996686935425,
      "logps/chosen": -35588.2109375,
      "logps/pi_response": -35588.0078125,
      "logps/ref_response": -1048.0765380859375,
      "logps/rejected": -35588.2109375,
      "loss": 0.6931,
      "rewards/accuracies": 0.0,
      "rewards/chosen": -345.4007263183594,
      "rewards/margins": 0.0,
      "rewards/rejected": -345.4007263183594,
      "step": 50
    },
    {
      "epoch": 0.9874476987447699,
      "step": 59,
      "total_flos": 0.0,
      "train_loss": 0.6931423080169549,
      "train_runtime": 1323.6379,
      "train_samples_per_second": 11.546,
      "train_steps_per_second": 0.045
    }
  ],
  "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": 4,
  "trial_name": null,
  "trial_params": null
}