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": 78.60790491305158,
      "learning_rate": 8.333333333333333e-08,
      "logits/chosen": -1.0250325202941895,
      "logits/rejected": -0.7656487226486206,
      "logps/chosen": -238.23211669921875,
      "logps/pi_response": -67.60678100585938,
      "logps/ref_response": -67.60678100585938,
      "logps/rejected": -473.57257080078125,
      "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": 47.26665166784664,
      "learning_rate": 4.930057285201027e-07,
      "logits/chosen": -0.7590005397796631,
      "logits/rejected": -0.534713625907898,
      "logps/chosen": -269.3804931640625,
      "logps/pi_response": -86.47183227539062,
      "logps/ref_response": -85.30850982666016,
      "logps/rejected": -577.9660034179688,
      "loss": 0.6322,
      "rewards/accuracies": 0.6284722089767456,
      "rewards/chosen": -0.17737312614917755,
      "rewards/margins": 0.3175060451030731,
      "rewards/rejected": -0.49487918615341187,
      "step": 10
    },
    {
      "epoch": 0.33472803347280333,
      "grad_norm": 78.10720741885179,
      "learning_rate": 4.187457503795526e-07,
      "logits/chosen": 0.7693876624107361,
      "logits/rejected": 1.1501473188400269,
      "logps/chosen": -397.8346862792969,
      "logps/pi_response": -133.69357299804688,
      "logps/ref_response": -84.12879943847656,
      "logps/rejected": -852.2254028320312,
      "loss": 0.6126,
      "rewards/accuracies": 0.778124988079071,
      "rewards/chosen": -1.3858518600463867,
      "rewards/margins": 1.7604806423187256,
      "rewards/rejected": -3.1463327407836914,
      "step": 20
    },
    {
      "epoch": 0.502092050209205,
      "grad_norm": 44.300549506110464,
      "learning_rate": 2.8691164100062034e-07,
      "logits/chosen": 1.0299599170684814,
      "logits/rejected": 1.5359565019607544,
      "logps/chosen": -398.71575927734375,
      "logps/pi_response": -156.10476684570312,
      "logps/ref_response": -84.0820541381836,
      "logps/rejected": -803.1231689453125,
      "loss": 0.4797,
      "rewards/accuracies": 0.731249988079071,
      "rewards/chosen": -1.3773680925369263,
      "rewards/margins": 1.4377108812332153,
      "rewards/rejected": -2.8150792121887207,
      "step": 30
    },
    {
      "epoch": 0.6694560669456067,
      "grad_norm": 40.9925395869338,
      "learning_rate": 1.4248369943086995e-07,
      "logits/chosen": 0.7738146781921387,
      "logits/rejected": 1.347411036491394,
      "logps/chosen": -327.626953125,
      "logps/pi_response": -121.0374526977539,
      "logps/ref_response": -77.38616943359375,
      "logps/rejected": -800.9747314453125,
      "loss": 0.4483,
      "rewards/accuracies": 0.7718750238418579,
      "rewards/chosen": -0.895713210105896,
      "rewards/margins": 1.4477618932724,
      "rewards/rejected": -2.343475103378296,
      "step": 40
    },
    {
      "epoch": 0.8368200836820083,
      "grad_norm": 43.60271544881446,
      "learning_rate": 3.473909705816111e-08,
      "logits/chosen": 0.9320430755615234,
      "logits/rejected": 1.5464847087860107,
      "logps/chosen": -350.8163757324219,
      "logps/pi_response": -132.66949462890625,
      "logps/ref_response": -84.61216735839844,
      "logps/rejected": -686.16015625,
      "loss": 0.4596,
      "rewards/accuracies": 0.7406250238418579,
      "rewards/chosen": -0.8762038350105286,
      "rewards/margins": 1.1485484838485718,
      "rewards/rejected": -2.024752378463745,
      "step": 50
    },
    {
      "epoch": 0.9874476987447699,
      "step": 59,
      "total_flos": 0.0,
      "train_loss": 0.5195278232380495,
      "train_runtime": 2549.7708,
      "train_samples_per_second": 5.994,
      "train_steps_per_second": 0.023
    }
  ],
  "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
}