trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.8363636363636364,
  "eval_steps": 40,
  "global_step": 39,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.36363636363636365,
      "grad_norm": 109.26669040334116,
      "learning_rate": 5e-07,
      "logits/chosen": -2.7417855262756348,
      "logits/rejected": -2.731990098953247,
      "logps/chosen": -402.4177551269531,
      "logps/rejected": -259.22088623046875,
      "loss": 0.6813,
      "rewards/accuracies": 0.3687500059604645,
      "rewards/chosen": 0.03968711942434311,
      "rewards/margins": 0.023939523845911026,
      "rewards/rejected": 0.015747597441077232,
      "step": 5
    },
    {
      "epoch": 0.7272727272727273,
      "grad_norm": 54.44026233316852,
      "learning_rate": 1e-06,
      "logits/chosen": -2.695871353149414,
      "logits/rejected": -2.6486518383026123,
      "logps/chosen": -395.3858947753906,
      "logps/rejected": -254.0742645263672,
      "loss": 0.4811,
      "rewards/accuracies": 0.8125,
      "rewards/chosen": 1.3482853174209595,
      "rewards/margins": 0.7143442630767822,
      "rewards/rejected": 0.6339409947395325,
      "step": 10
    },
    {
      "epoch": 1.0909090909090908,
      "grad_norm": 40.622864628442045,
      "learning_rate": 9.284285880837946e-07,
      "logits/chosen": -2.4416801929473877,
      "logits/rejected": -2.394608736038208,
      "logps/chosen": -397.1449890136719,
      "logps/rejected": -282.96759033203125,
      "loss": 0.3285,
      "rewards/accuracies": 0.856249988079071,
      "rewards/chosen": 4.510091781616211,
      "rewards/margins": 2.8856072425842285,
      "rewards/rejected": 1.6244843006134033,
      "step": 15
    },
    {
      "epoch": 1.4545454545454546,
      "grad_norm": 16.90254803263314,
      "learning_rate": 7.342042203498951e-07,
      "logits/chosen": -2.289841413497925,
      "logits/rejected": -2.265059232711792,
      "logps/chosen": -343.917724609375,
      "logps/rejected": -253.5391082763672,
      "loss": 0.1267,
      "rewards/accuracies": 0.9312499761581421,
      "rewards/chosen": 5.056896686553955,
      "rewards/margins": 4.160580158233643,
      "rewards/rejected": 0.896316647529602,
      "step": 20
    },
    {
      "epoch": 1.8181818181818183,
      "grad_norm": 29.584850630021336,
      "learning_rate": 4.7293054570729126e-07,
      "logits/chosen": -2.163829803466797,
      "logits/rejected": -2.1356260776519775,
      "logps/chosen": -357.2837829589844,
      "logps/rejected": -278.4024963378906,
      "loss": 0.093,
      "rewards/accuracies": 0.9624999761581421,
      "rewards/chosen": 5.750279426574707,
      "rewards/margins": 5.663321495056152,
      "rewards/rejected": 0.08695853501558304,
      "step": 25
    },
    {
      "epoch": 2.1818181818181817,
      "grad_norm": 7.589156329261643,
      "learning_rate": 2.1940646731880885e-07,
      "logits/chosen": -2.1467955112457275,
      "logits/rejected": -2.1011250019073486,
      "logps/chosen": -361.9977111816406,
      "logps/rejected": -266.29840087890625,
      "loss": 0.0614,
      "rewards/accuracies": 0.981249988079071,
      "rewards/chosen": 5.7993059158325195,
      "rewards/margins": 5.681790351867676,
      "rewards/rejected": 0.11751584708690643,
      "step": 30
    },
    {
      "epoch": 2.5454545454545454,
      "grad_norm": 4.596558322612688,
      "learning_rate": 4.621229016452155e-08,
      "logits/chosen": -2.1508798599243164,
      "logits/rejected": -2.1145217418670654,
      "logps/chosen": -361.5357360839844,
      "logps/rejected": -262.2801208496094,
      "loss": 0.0241,
      "rewards/accuracies": 0.9937499761581421,
      "rewards/chosen": 6.36221981048584,
      "rewards/margins": 6.471512794494629,
      "rewards/rejected": -0.10929267108440399,
      "step": 35
    },
    {
      "epoch": 2.8363636363636364,
      "step": 39,
      "total_flos": 459455468666880.0,
      "train_loss": 0.2336020725659835,
      "train_runtime": 1351.9222,
      "train_samples_per_second": 7.685,
      "train_steps_per_second": 0.029
    }
  ],
  "logging_steps": 5,
  "max_steps": 39,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 40,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 459455468666880.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}