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": 53.25663135243593,
      "learning_rate": 8.333333333333333e-08,
      "logits/chosen": -2.60155987739563,
      "logits/rejected": -2.560150146484375,
      "logps/chosen": -302.1136474609375,
      "logps/pi_response": -142.7164764404297,
      "logps/ref_response": -142.7164764404297,
      "logps/rejected": -372.4461364746094,
      "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": 33.71802345048987,
      "learning_rate": 4.930057285201027e-07,
      "logits/chosen": -2.5734050273895264,
      "logits/rejected": -2.528841257095337,
      "logps/chosen": -264.94677734375,
      "logps/pi_response": -133.82980346679688,
      "logps/ref_response": -121.70560455322266,
      "logps/rejected": -426.4489440917969,
      "loss": 0.6615,
      "rewards/accuracies": 0.5694444179534912,
      "rewards/chosen": -0.24433700740337372,
      "rewards/margins": 0.1003943383693695,
      "rewards/rejected": -0.34473133087158203,
      "step": 10
    },
    {
      "epoch": 0.33472803347280333,
      "grad_norm": 42.366900514115656,
      "learning_rate": 4.187457503795526e-07,
      "logits/chosen": -2.4807186126708984,
      "logits/rejected": -2.4149527549743652,
      "logps/chosen": -322.50579833984375,
      "logps/pi_response": -146.48080444335938,
      "logps/ref_response": -126.39298248291016,
      "logps/rejected": -567.7706298828125,
      "loss": 0.5783,
      "rewards/accuracies": 0.731249988079071,
      "rewards/chosen": -0.8342545628547668,
      "rewards/margins": 0.9264798164367676,
      "rewards/rejected": -1.7607343196868896,
      "step": 20
    },
    {
      "epoch": 0.502092050209205,
      "grad_norm": 37.21235124072302,
      "learning_rate": 2.8691164100062034e-07,
      "logits/chosen": -2.3930041790008545,
      "logits/rejected": -2.3236992359161377,
      "logps/chosen": -341.8916320800781,
      "logps/pi_response": -139.42262268066406,
      "logps/ref_response": -122.14532470703125,
      "logps/rejected": -547.8569946289062,
      "loss": 0.5214,
      "rewards/accuracies": 0.746874988079071,
      "rewards/chosen": -0.8386001586914062,
      "rewards/margins": 0.8395956158638,
      "rewards/rejected": -1.6781957149505615,
      "step": 30
    },
    {
      "epoch": 0.6694560669456067,
      "grad_norm": 37.34482955607721,
      "learning_rate": 1.4248369943086995e-07,
      "logits/chosen": -2.4054012298583984,
      "logits/rejected": -2.3128952980041504,
      "logps/chosen": -299.69805908203125,
      "logps/pi_response": -133.9350128173828,
      "logps/ref_response": -125.8031234741211,
      "logps/rejected": -555.8270263671875,
      "loss": 0.4991,
      "rewards/accuracies": 0.778124988079071,
      "rewards/chosen": -0.5645470023155212,
      "rewards/margins": 0.8838322758674622,
      "rewards/rejected": -1.4483792781829834,
      "step": 40
    },
    {
      "epoch": 0.8368200836820083,
      "grad_norm": 25.861200618216518,
      "learning_rate": 3.473909705816111e-08,
      "logits/chosen": -2.3677544593811035,
      "logits/rejected": -2.3059608936309814,
      "logps/chosen": -317.23077392578125,
      "logps/pi_response": -134.0728302001953,
      "logps/ref_response": -123.98958587646484,
      "logps/rejected": -542.5103759765625,
      "loss": 0.5018,
      "rewards/accuracies": 0.7593749761581421,
      "rewards/chosen": -0.656657874584198,
      "rewards/margins": 0.8108199238777161,
      "rewards/rejected": -1.467477798461914,
      "step": 50
    },
    {
      "epoch": 0.9874476987447699,
      "step": 59,
      "total_flos": 0.0,
      "train_loss": 0.5438056961964752,
      "train_runtime": 2605.9239,
      "train_samples_per_second": 5.865,
      "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
}