trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.986666666666667,
  "eval_steps": 500,
  "global_step": 84,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.35555555555555557,
      "grad_norm": 8.846159934997559,
      "learning_rate": 4.997807075247147e-06,
      "logits/chosen": -0.43419259786605835,
      "logits/rejected": -0.5061447024345398,
      "logps/chosen": -1.7686179876327515,
      "logps/rejected": -2.1890695095062256,
      "loss": 1.8236,
      "odds_ratio_loss": 15.943506240844727,
      "rewards/accuracies": 0.762499988079071,
      "rewards/chosen": -0.17686179280281067,
      "rewards/margins": 0.042045146226882935,
      "rewards/rejected": -0.2189069539308548,
      "sft_loss": 0.22924074530601501,
      "step": 10
    },
    {
      "epoch": 0.7111111111111111,
      "grad_norm": 6.154786586761475,
      "learning_rate": 4.7392794005985324e-06,
      "logits/chosen": -0.36880916357040405,
      "logits/rejected": -0.45593205094337463,
      "logps/chosen": -1.6094945669174194,
      "logps/rejected": -2.0511698722839355,
      "loss": 1.6628,
      "odds_ratio_loss": 14.580706596374512,
      "rewards/accuracies": 0.731249988079071,
      "rewards/chosen": -0.1609494537115097,
      "rewards/margins": 0.04416752606630325,
      "rewards/rejected": -0.20511698722839355,
      "sft_loss": 0.20470395684242249,
      "step": 20
    },
    {
      "epoch": 1.0666666666666667,
      "grad_norm": 5.7998881340026855,
      "learning_rate": 4.093559974371725e-06,
      "logits/chosen": -0.41758084297180176,
      "logits/rejected": -0.5090541839599609,
      "logps/chosen": -1.4586153030395508,
      "logps/rejected": -1.9464032649993896,
      "loss": 1.5084,
      "odds_ratio_loss": 13.208415031433105,
      "rewards/accuracies": 0.8187500238418579,
      "rewards/chosen": -0.14586155116558075,
      "rewards/margins": 0.04877878352999687,
      "rewards/rejected": -0.19464033842086792,
      "sft_loss": 0.18757374584674835,
      "step": 30
    },
    {
      "epoch": 1.4222222222222223,
      "grad_norm": 6.195462703704834,
      "learning_rate": 3.1722995515381644e-06,
      "logits/chosen": -0.4006771445274353,
      "logits/rejected": -0.47693929076194763,
      "logps/chosen": -1.2410690784454346,
      "logps/rejected": -1.7496509552001953,
      "loss": 1.2886,
      "odds_ratio_loss": 11.358288764953613,
      "rewards/accuracies": 0.8062499761581421,
      "rewards/chosen": -0.12410689890384674,
      "rewards/margins": 0.05085818096995354,
      "rewards/rejected": -0.17496509850025177,
      "sft_loss": 0.1527349054813385,
      "step": 40
    },
    {
      "epoch": 1.7777777777777777,
      "grad_norm": 5.72018575668335,
      "learning_rate": 2.134792428593971e-06,
      "logits/chosen": -0.40342792868614197,
      "logits/rejected": -0.47864627838134766,
      "logps/chosen": -1.026789903640747,
      "logps/rejected": -1.4996291399002075,
      "loss": 1.0765,
      "odds_ratio_loss": 9.527546882629395,
      "rewards/accuracies": 0.7875000238418579,
      "rewards/chosen": -0.10267897695302963,
      "rewards/margins": 0.047283925116062164,
      "rewards/rejected": -0.14996293187141418,
      "sft_loss": 0.12372473627328873,
      "step": 50
    },
    {
      "epoch": 2.1333333333333333,
      "grad_norm": 4.517279624938965,
      "learning_rate": 1.160433012552508e-06,
      "logits/chosen": -0.4058423936367035,
      "logits/rejected": -0.49612194299697876,
      "logps/chosen": -0.8940722346305847,
      "logps/rejected": -1.3635480403900146,
      "loss": 0.9456,
      "odds_ratio_loss": 8.374347686767578,
      "rewards/accuracies": 0.7749999761581421,
      "rewards/chosen": -0.08940722793340683,
      "rewards/margins": 0.046947579830884933,
      "rewards/rejected": -0.13635481894016266,
      "sft_loss": 0.10813729465007782,
      "step": 60
    },
    {
      "epoch": 2.488888888888889,
      "grad_norm": 3.7044575214385986,
      "learning_rate": 4.1769689822475147e-07,
      "logits/chosen": -0.41305217146873474,
      "logits/rejected": -0.4884258210659027,
      "logps/chosen": -0.8132768869400024,
      "logps/rejected": -1.3795119524002075,
      "loss": 0.8557,
      "odds_ratio_loss": 7.654118537902832,
      "rewards/accuracies": 0.824999988079071,
      "rewards/chosen": -0.08132769167423248,
      "rewards/margins": 0.05662349984049797,
      "rewards/rejected": -0.13795119524002075,
      "sft_loss": 0.09026216715574265,
      "step": 70
    },
    {
      "epoch": 2.8444444444444446,
      "grad_norm": 3.704728603363037,
      "learning_rate": 3.5009907323737826e-08,
      "logits/chosen": -0.4267275929450989,
      "logits/rejected": -0.5049682855606079,
      "logps/chosen": -0.7886781096458435,
      "logps/rejected": -1.2518656253814697,
      "loss": 0.8364,
      "odds_ratio_loss": 7.4754438400268555,
      "rewards/accuracies": 0.768750011920929,
      "rewards/chosen": -0.07886780798435211,
      "rewards/margins": 0.046318747103214264,
      "rewards/rejected": -0.12518656253814697,
      "sft_loss": 0.08889047056436539,
      "step": 80
    },
    {
      "epoch": 2.986666666666667,
      "step": 84,
      "total_flos": 1.516498186272768e+16,
      "train_loss": 1.23334006468455,
      "train_runtime": 194.0073,
      "train_samples_per_second": 6.959,
      "train_steps_per_second": 0.433
    }
  ],
  "logging_steps": 10,
  "max_steps": 84,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 1.516498186272768e+16,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}