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": 11.933876924387047,
      "learning_rate": 8.333333333333333e-08,
      "logits/chosen": -2.8434743881225586,
      "logits/rejected": -2.769583225250244,
      "logps/chosen": -178.4530487060547,
      "logps/pi_response": -97.4625244140625,
      "logps/ref_response": -97.4625244140625,
      "logps/rejected": -233.52891540527344,
      "loss": 0.6907,
      "rewards/accuracies": 0.0,
      "rewards/chosen": 0.0,
      "rewards/margins": 0.0,
      "rewards/rejected": 0.0,
      "step": 1
    },
    {
      "epoch": 0.16736401673640167,
      "grad_norm": 11.70347902906233,
      "learning_rate": 4.930057285201027e-07,
      "logits/chosen": -2.754892349243164,
      "logits/rejected": -2.7298169136047363,
      "logps/chosen": -225.98452758789062,
      "logps/pi_response": -120.69906616210938,
      "logps/ref_response": -119.74172973632812,
      "logps/rejected": -266.0158996582031,
      "loss": 0.6824,
      "rewards/accuracies": 0.5972222089767456,
      "rewards/chosen": -0.03290151432156563,
      "rewards/margins": 0.02144717425107956,
      "rewards/rejected": -0.054348692297935486,
      "step": 10
    },
    {
      "epoch": 0.33472803347280333,
      "grad_norm": 25.42242304108054,
      "learning_rate": 4.187457503795526e-07,
      "logits/chosen": -2.799253225326538,
      "logits/rejected": -2.749119281768799,
      "logps/chosen": -259.5742492675781,
      "logps/pi_response": -116.0076904296875,
      "logps/ref_response": -118.667724609375,
      "logps/rejected": -312.127685546875,
      "loss": 0.6307,
      "rewards/accuracies": 0.643750011920929,
      "rewards/chosen": -0.2481091022491455,
      "rewards/margins": 0.2937210500240326,
      "rewards/rejected": -0.5418301820755005,
      "step": 20
    },
    {
      "epoch": 0.502092050209205,
      "grad_norm": 15.819520353220648,
      "learning_rate": 2.8691164100062034e-07,
      "logits/chosen": -2.7805113792419434,
      "logits/rejected": -2.7487761974334717,
      "logps/chosen": -287.7320251464844,
      "logps/pi_response": -147.50924682617188,
      "logps/ref_response": -124.9267807006836,
      "logps/rejected": -356.9125061035156,
      "loss": 0.5822,
      "rewards/accuracies": 0.706250011920929,
      "rewards/chosen": -0.3946269750595093,
      "rewards/margins": 0.5142250657081604,
      "rewards/rejected": -0.9088519811630249,
      "step": 30
    },
    {
      "epoch": 0.6694560669456067,
      "grad_norm": 15.495549373699717,
      "learning_rate": 1.4248369943086995e-07,
      "logits/chosen": -2.677691698074341,
      "logits/rejected": -2.636936902999878,
      "logps/chosen": -273.7499694824219,
      "logps/pi_response": -152.98040771484375,
      "logps/ref_response": -115.91612243652344,
      "logps/rejected": -367.84100341796875,
      "loss": 0.5526,
      "rewards/accuracies": 0.6937500238418579,
      "rewards/chosen": -0.594318687915802,
      "rewards/margins": 0.5897426605224609,
      "rewards/rejected": -1.1840614080429077,
      "step": 40
    },
    {
      "epoch": 0.8368200836820083,
      "grad_norm": 20.343668088012315,
      "learning_rate": 3.473909705816111e-08,
      "logits/chosen": -2.7078301906585693,
      "logits/rejected": -2.6557488441467285,
      "logps/chosen": -264.0274353027344,
      "logps/pi_response": -144.65487670898438,
      "logps/ref_response": -106.76103210449219,
      "logps/rejected": -335.317626953125,
      "loss": 0.5628,
      "rewards/accuracies": 0.6937500238418579,
      "rewards/chosen": -0.5633394122123718,
      "rewards/margins": 0.49273037910461426,
      "rewards/rejected": -1.0560699701309204,
      "step": 50
    },
    {
      "epoch": 0.9874476987447699,
      "step": 59,
      "total_flos": 0.0,
      "train_loss": 0.5970514467207052,
      "train_runtime": 1330.0906,
      "train_samples_per_second": 11.49,
      "train_steps_per_second": 0.044
    }
  ],
  "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
}