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": 12.20743074581383,
      "learning_rate": 8.333333333333333e-08,
      "logits/chosen": -2.825530767440796,
      "logits/rejected": -2.756951332092285,
      "logps/chosen": -210.2147979736328,
      "logps/pi_response": -115.76622009277344,
      "logps/ref_response": -115.76622009277344,
      "logps/rejected": -234.01425170898438,
      "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": 12.223946579272333,
      "learning_rate": 4.930057285201027e-07,
      "logits/chosen": -2.7576725482940674,
      "logits/rejected": -2.7086384296417236,
      "logps/chosen": -221.15676879882812,
      "logps/pi_response": -116.39059448242188,
      "logps/ref_response": -115.38680267333984,
      "logps/rejected": -261.13031005859375,
      "loss": 0.6862,
      "rewards/accuracies": 0.5416666865348816,
      "rewards/chosen": -0.04135013371706009,
      "rewards/margins": 0.019138270989060402,
      "rewards/rejected": -0.06048841029405594,
      "step": 10
    },
    {
      "epoch": 0.33472803347280333,
      "grad_norm": 27.23026162056322,
      "learning_rate": 4.187457503795526e-07,
      "logits/chosen": -2.7793827056884766,
      "logits/rejected": -2.7369754314422607,
      "logps/chosen": -245.52627563476562,
      "logps/pi_response": -110.24317932128906,
      "logps/ref_response": -113.51313781738281,
      "logps/rejected": -298.79180908203125,
      "loss": 0.6346,
      "rewards/accuracies": 0.6468750238418579,
      "rewards/chosen": -0.2135976105928421,
      "rewards/margins": 0.2609703838825226,
      "rewards/rejected": -0.4745679795742035,
      "step": 20
    },
    {
      "epoch": 0.502092050209205,
      "grad_norm": 16.564474644022212,
      "learning_rate": 2.8691164100062034e-07,
      "logits/chosen": -2.765554428100586,
      "logits/rejected": -2.7183127403259277,
      "logps/chosen": -278.768798828125,
      "logps/pi_response": -138.74417114257812,
      "logps/ref_response": -120.3040542602539,
      "logps/rejected": -346.51177978515625,
      "loss": 0.5838,
      "rewards/accuracies": 0.690625011920929,
      "rewards/chosen": -0.47783130407333374,
      "rewards/margins": 0.4638592302799225,
      "rewards/rejected": -0.9416905641555786,
      "step": 30
    },
    {
      "epoch": 0.6694560669456067,
      "grad_norm": 16.782748799564995,
      "learning_rate": 1.4248369943086995e-07,
      "logits/chosen": -2.691978693008423,
      "logits/rejected": -2.649599552154541,
      "logps/chosen": -283.6568298339844,
      "logps/pi_response": -144.93222045898438,
      "logps/ref_response": -112.8275375366211,
      "logps/rejected": -357.76116943359375,
      "loss": 0.5652,
      "rewards/accuracies": 0.6812499761581421,
      "rewards/chosen": -0.563094437122345,
      "rewards/margins": 0.47530126571655273,
      "rewards/rejected": -1.0383957624435425,
      "step": 40
    },
    {
      "epoch": 0.8368200836820083,
      "grad_norm": 17.23293020250368,
      "learning_rate": 3.473909705816111e-08,
      "logits/chosen": -2.67022705078125,
      "logits/rejected": -2.62345290184021,
      "logps/chosen": -279.62060546875,
      "logps/pi_response": -149.0137481689453,
      "logps/ref_response": -111.8585433959961,
      "logps/rejected": -333.55364990234375,
      "loss": 0.5589,
      "rewards/accuracies": 0.699999988079071,
      "rewards/chosen": -0.5371103882789612,
      "rewards/margins": 0.4536103308200836,
      "rewards/rejected": -0.9907207489013672,
      "step": 50
    },
    {
      "epoch": 0.9874476987447699,
      "step": 59,
      "total_flos": 0.0,
      "train_loss": 0.5968084577786721,
      "train_runtime": 2657.7029,
      "train_samples_per_second": 5.75,
      "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
}