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": 53.776747896947,
      "learning_rate": 8.333333333333333e-08,
      "logits/chosen": -2.794323682785034,
      "logits/rejected": -2.8258514404296875,
      "logps/chosen": -210.50025939941406,
      "logps/pi_response": -163.35888671875,
      "logps/ref_response": -163.35888671875,
      "logps/rejected": -543.4459228515625,
      "loss": 0.6892,
      "rewards/accuracies": 0.0,
      "rewards/chosen": 0.0,
      "rewards/margins": 0.0,
      "rewards/rejected": 0.0,
      "step": 1
    },
    {
      "epoch": 0.16736401673640167,
      "grad_norm": 48.213055106513124,
      "learning_rate": 4.930057285201027e-07,
      "logits/chosen": -2.633598804473877,
      "logits/rejected": -2.5702714920043945,
      "logps/chosen": -271.011474609375,
      "logps/pi_response": -141.91607666015625,
      "logps/ref_response": -125.00880432128906,
      "logps/rejected": -440.8132629394531,
      "loss": 0.6509,
      "rewards/accuracies": 0.6527777910232544,
      "rewards/chosen": -0.3297349214553833,
      "rewards/margins": 0.10624222457408905,
      "rewards/rejected": -0.43597719073295593,
      "step": 10
    },
    {
      "epoch": 0.33472803347280333,
      "grad_norm": 38.453358160886886,
      "learning_rate": 4.187457503795526e-07,
      "logits/chosen": -2.5579605102539062,
      "logits/rejected": -2.492035388946533,
      "logps/chosen": -311.72369384765625,
      "logps/pi_response": -150.08323669433594,
      "logps/ref_response": -116.40202331542969,
      "logps/rejected": -561.711181640625,
      "loss": 0.5785,
      "rewards/accuracies": 0.75,
      "rewards/chosen": -0.9310864210128784,
      "rewards/margins": 1.0276349782943726,
      "rewards/rejected": -1.9587215185165405,
      "step": 20
    },
    {
      "epoch": 0.502092050209205,
      "grad_norm": 25.46935093592709,
      "learning_rate": 2.8691164100062034e-07,
      "logits/chosen": -2.4656379222869873,
      "logits/rejected": -2.4074764251708984,
      "logps/chosen": -325.81707763671875,
      "logps/pi_response": -152.36343383789062,
      "logps/ref_response": -120.72613525390625,
      "logps/rejected": -509.164306640625,
      "loss": 0.5058,
      "rewards/accuracies": 0.7562500238418579,
      "rewards/chosen": -0.8586891889572144,
      "rewards/margins": 0.7089561223983765,
      "rewards/rejected": -1.5676453113555908,
      "step": 30
    },
    {
      "epoch": 0.6694560669456067,
      "grad_norm": 30.91533852830777,
      "learning_rate": 1.4248369943086995e-07,
      "logits/chosen": -2.4529154300689697,
      "logits/rejected": -2.33906626701355,
      "logps/chosen": -340.20733642578125,
      "logps/pi_response": -151.04525756835938,
      "logps/ref_response": -117.29234313964844,
      "logps/rejected": -557.9139404296875,
      "loss": 0.4974,
      "rewards/accuracies": 0.793749988079071,
      "rewards/chosen": -0.8393608331680298,
      "rewards/margins": 0.8147125244140625,
      "rewards/rejected": -1.6540733575820923,
      "step": 40
    },
    {
      "epoch": 0.8368200836820083,
      "grad_norm": 25.998689260778725,
      "learning_rate": 3.473909705816111e-08,
      "logits/chosen": -2.4658045768737793,
      "logits/rejected": -2.375366687774658,
      "logps/chosen": -308.40985107421875,
      "logps/pi_response": -153.0810546875,
      "logps/ref_response": -127.05342102050781,
      "logps/rejected": -543.6156005859375,
      "loss": 0.4801,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": -0.7126681208610535,
      "rewards/margins": 0.9508953094482422,
      "rewards/rejected": -1.6635633707046509,
      "step": 50
    },
    {
      "epoch": 0.9874476987447699,
      "step": 59,
      "total_flos": 0.0,
      "train_loss": 0.5329152315349902,
      "train_runtime": 1310.6094,
      "train_samples_per_second": 11.661,
      "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
}