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": 179.34174359528225,
      "learning_rate": 1.6666666666666664e-08,
      "logits/chosen": 2.640702247619629,
      "logits/rejected": 2.6418023109436035,
      "logps/chosen": -476.8863830566406,
      "logps/pi_response": -236.9886474609375,
      "logps/ref_response": -236.9886474609375,
      "logps/rejected": -646.126708984375,
      "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": 172.47699668231658,
      "learning_rate": 9.860114570402053e-08,
      "logits/chosen": 2.442354440689087,
      "logits/rejected": 2.56209135055542,
      "logps/chosen": -468.1467590332031,
      "logps/pi_response": -212.0287322998047,
      "logps/ref_response": -207.71839904785156,
      "logps/rejected": -681.3190307617188,
      "loss": 0.6718,
      "rewards/accuracies": 0.6076388955116272,
      "rewards/chosen": -0.020253373309969902,
      "rewards/margins": 0.04976692795753479,
      "rewards/rejected": -0.07002030313014984,
      "step": 10
    },
    {
      "epoch": 0.33472803347280333,
      "grad_norm": 80.95712523741385,
      "learning_rate": 8.374915007591053e-08,
      "logits/chosen": 2.9408648014068604,
      "logits/rejected": 2.794188976287842,
      "logps/chosen": -483.8638610839844,
      "logps/pi_response": -394.9246826171875,
      "logps/ref_response": -259.81427001953125,
      "logps/rejected": -825.8529052734375,
      "loss": 0.4733,
      "rewards/accuracies": 0.8531249761581421,
      "rewards/chosen": -0.3374873697757721,
      "rewards/margins": 1.2408525943756104,
      "rewards/rejected": -1.57833993434906,
      "step": 20
    },
    {
      "epoch": 0.502092050209205,
      "grad_norm": 65.33777634592894,
      "learning_rate": 5.738232820012406e-08,
      "logits/chosen": 2.827476978302002,
      "logits/rejected": 3.1196391582489014,
      "logps/chosen": -523.5798950195312,
      "logps/pi_response": -502.6634826660156,
      "logps/ref_response": -225.60720825195312,
      "logps/rejected": -1013.6014404296875,
      "loss": 0.3726,
      "rewards/accuracies": 0.831250011920929,
      "rewards/chosen": -0.7283425331115723,
      "rewards/margins": 2.394611358642578,
      "rewards/rejected": -3.1229538917541504,
      "step": 30
    },
    {
      "epoch": 0.6694560669456067,
      "grad_norm": 85.8232407562363,
      "learning_rate": 2.8496739886173992e-08,
      "logits/chosen": 3.10669207572937,
      "logits/rejected": 3.293214797973633,
      "logps/chosen": -575.4429931640625,
      "logps/pi_response": -713.3890380859375,
      "logps/ref_response": -235.89822387695312,
      "logps/rejected": -1242.5897216796875,
      "loss": 0.3664,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -1.0589369535446167,
      "rewards/margins": 4.259150505065918,
      "rewards/rejected": -5.318087100982666,
      "step": 40
    },
    {
      "epoch": 0.8368200836820083,
      "grad_norm": 92.81078080419626,
      "learning_rate": 6.947819411632222e-09,
      "logits/chosen": 3.1707630157470703,
      "logits/rejected": 3.55635142326355,
      "logps/chosen": -652.9293212890625,
      "logps/pi_response": -746.1058959960938,
      "logps/ref_response": -231.4905242919922,
      "logps/rejected": -1275.537841796875,
      "loss": 0.3613,
      "rewards/accuracies": 0.8218749761581421,
      "rewards/chosen": -1.3429057598114014,
      "rewards/margins": 4.283051490783691,
      "rewards/rejected": -5.625957012176514,
      "step": 50
    },
    {
      "epoch": 0.9874476987447699,
      "step": 59,
      "total_flos": 0.0,
      "train_loss": 0.43783688141127763,
      "train_runtime": 2738.5584,
      "train_samples_per_second": 5.581,
      "train_steps_per_second": 0.022
    }
  ],
  "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
}