{
  "best_metric": 0.3008906841278076,
  "best_model_checkpoint": "/datastor1/fcyin/finetuned_checkpoints/truthfulqa/llama2_7B_truthfulqa_lofit_seed42_96heads/checkpoint-123",
  "epoch": 5.0,
  "eval_steps": 500,
  "global_step": 205,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 0.6008107662200928,
      "learning_rate": 0.004,
      "logits/chosen": -2.5561070442199707,
      "logits/rejected": -2.650902271270752,
      "logps/chosen": -29.893611907958984,
      "logps/rejected": -31.828697204589844,
      "loss": 0.2945,
      "rewards/accuracies": 0.8597561120986938,
      "rewards/chosen": -0.5662117004394531,
      "rewards/margins": 2.810072422027588,
      "rewards/rejected": -3.376283884048462,
      "step": 41
    },
    {
      "epoch": 1.0,
      "eval_logits/chosen": -3.046513319015503,
      "eval_logits/rejected": -3.213273048400879,
      "eval_logps/chosen": -30.533157348632812,
      "eval_logps/rejected": -33.57004165649414,
      "eval_loss": 0.6931471228599548,
      "eval_rewards/accuracies": 0.0,
      "eval_rewards/chosen": 0.0,
      "eval_rewards/margins": 0.0,
      "eval_rewards/rejected": 0.0,
      "eval_runtime": 7.6196,
      "eval_samples_per_second": 10.762,
      "eval_steps_per_second": 0.394,
      "step": 41
    },
    {
      "epoch": 2.0,
      "grad_norm": 0.98629230260849,
      "learning_rate": 0.003,
      "logits/chosen": -2.6476099491119385,
      "logits/rejected": -2.767660140991211,
      "logps/chosen": -29.75066375732422,
      "logps/rejected": -38.1750602722168,
      "loss": 0.0504,
      "rewards/accuracies": 0.9969512224197388,
      "rewards/chosen": -0.5129372477531433,
      "rewards/margins": 6.039527893066406,
      "rewards/rejected": -6.552464962005615,
      "step": 82
    },
    {
      "epoch": 2.0,
      "eval_logits/chosen": -2.8705577850341797,
      "eval_logits/rejected": -3.025930166244507,
      "eval_logps/chosen": -30.89835548400879,
      "eval_logps/rejected": -37.61328887939453,
      "eval_loss": 0.3324950039386749,
      "eval_rewards/accuracies": 0.8530092239379883,
      "eval_rewards/chosen": -0.18259893357753754,
      "eval_rewards/margins": 1.8390239477157593,
      "eval_rewards/rejected": -2.021622896194458,
      "eval_runtime": 7.4476,
      "eval_samples_per_second": 11.01,
      "eval_steps_per_second": 0.403,
      "step": 82
    },
    {
      "epoch": 3.0,
      "grad_norm": 2.8237621784210205,
      "learning_rate": 0.002,
      "logits/chosen": -2.4965312480926514,
      "logits/rejected": -2.600482940673828,
      "logps/chosen": -30.458192825317383,
      "logps/rejected": -42.681915283203125,
      "loss": 0.0133,
      "rewards/accuracies": 1.0,
      "rewards/chosen": -0.8702911734580994,
      "rewards/margins": 7.938199520111084,
      "rewards/rejected": -8.808490753173828,
      "step": 123
    },
    {
      "epoch": 3.0,
      "eval_logits/chosen": -2.774709701538086,
      "eval_logits/rejected": -2.927616834640503,
      "eval_logps/chosen": -32.284488677978516,
      "eval_logps/rejected": -40.45302200317383,
      "eval_loss": 0.3008906841278076,
      "eval_rewards/accuracies": 0.8611111044883728,
      "eval_rewards/chosen": -0.875666081905365,
      "eval_rewards/margins": 2.5658228397369385,
      "eval_rewards/rejected": -3.4414889812469482,
      "eval_runtime": 7.444,
      "eval_samples_per_second": 11.016,
      "eval_steps_per_second": 0.403,
      "step": 123
    },
    {
      "epoch": 4.0,
      "grad_norm": 0.09415304660797119,
      "learning_rate": 0.001,
      "logits/chosen": -2.4607629776000977,
      "logits/rejected": -2.572495222091675,
      "logps/chosen": -30.935243606567383,
      "logps/rejected": -44.62324905395508,
      "loss": 0.0063,
      "rewards/accuracies": 1.0,
      "rewards/chosen": -1.088704228401184,
      "rewards/margins": 8.673280715942383,
      "rewards/rejected": -9.761984825134277,
      "step": 164
    },
    {
      "epoch": 4.0,
      "eval_logits/chosen": -2.737975835800171,
      "eval_logits/rejected": -2.888521432876587,
      "eval_logps/chosen": -32.865299224853516,
      "eval_logps/rejected": -41.451168060302734,
      "eval_loss": 0.3184911608695984,
      "eval_rewards/accuracies": 0.8611111044883728,
      "eval_rewards/chosen": -1.1660696268081665,
      "eval_rewards/margins": 2.774491310119629,
      "eval_rewards/rejected": -3.940561056137085,
      "eval_runtime": 7.5099,
      "eval_samples_per_second": 10.919,
      "eval_steps_per_second": 0.399,
      "step": 164
    },
    {
      "epoch": 5.0,
      "grad_norm": 0.0271140243858099,
      "learning_rate": 0.0,
      "logits/chosen": -2.4665253162384033,
      "logits/rejected": -2.5754239559173584,
      "logps/chosen": -31.26015853881836,
      "logps/rejected": -45.692237854003906,
      "loss": 0.0041,
      "rewards/accuracies": 1.0,
      "rewards/chosen": -1.2341755628585815,
      "rewards/margins": 9.071220397949219,
      "rewards/rejected": -10.305395126342773,
      "step": 205
    },
    {
      "epoch": 5.0,
      "eval_logits/chosen": -2.723881959915161,
      "eval_logits/rejected": -2.8738536834716797,
      "eval_logps/chosen": -33.00375747680664,
      "eval_logps/rejected": -41.7939453125,
      "eval_loss": 0.32048389315605164,
      "eval_rewards/accuracies": 0.8715277314186096,
      "eval_rewards/chosen": -1.2353010177612305,
      "eval_rewards/margins": 2.8766491413116455,
      "eval_rewards/rejected": -4.111950397491455,
      "eval_runtime": 7.4592,
      "eval_samples_per_second": 10.993,
      "eval_steps_per_second": 0.402,
      "step": 205
    }
  ],
  "logging_steps": 500,
  "max_steps": 205,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "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": 0.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}