{
  "best_global_step": 77,
  "best_metric": 0.6231179237365723,
  "best_model_checkpoint": "outputs/qwen35_dpo_ultralowmem_ref_free/checkpoint-77",
  "epoch": 1.0,
  "eval_steps": 200,
  "global_step": 77,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "entropy": 0.9454727958887815,
      "epoch": 0.13136288998357964,
      "grad_norm": 7.8125,
      "learning_rate": 4.919332367333748e-07,
      "logits/chosen": -1.8099737944940273,
      "logits/rejected": -1.832894790703956,
      "logps/chosen": -353.753306388855,
      "logps/rejected": -344.2661180496216,
      "loss": 0.6934457302093506,
      "mean_token_accuracy": 0.7538095749914646,
      "num_tokens": 1287126.0,
      "rewards/accuracies": 0.409375,
      "rewards/chosen": -0.00509019878518302,
      "rewards/margins": 0.0012955188169144094,
      "rewards/rejected": -0.006385717548255343,
      "step": 10
    },
    {
      "entropy": 0.960740290209651,
      "epoch": 0.2627257799671593,
      "grad_norm": 7.0625,
      "learning_rate": 4.4450893857960984e-07,
      "logits/chosen": -1.8059819363814396,
      "logits/rejected": -1.8327529228301944,
      "logps/chosen": -383.5576331138611,
      "logps/rejected": -348.3659210205078,
      "loss": 0.6661795139312744,
      "mean_token_accuracy": 0.7455122817307711,
      "num_tokens": 2584081.0,
      "rewards/accuracies": 0.696875,
      "rewards/chosen": 0.01022649770602584,
      "rewards/margins": 0.05804813946597278,
      "rewards/rejected": -0.04782164186508453,
      "step": 20
    },
    {
      "entropy": 0.9594713591039181,
      "epoch": 0.39408866995073893,
      "grad_norm": 6.34375,
      "learning_rate": 3.625509362044183e-07,
      "logits/chosen": -1.8090848916866544,
      "logits/rejected": -1.839708000919806,
      "logps/chosen": -367.3285415649414,
      "logps/rejected": -347.3939818382263,
      "loss": 0.6498698234558106,
      "mean_token_accuracy": 0.7490173149853945,
      "num_tokens": 3873129.0,
      "rewards/accuracies": 0.76875,
      "rewards/chosen": 0.01563858055451419,
      "rewards/margins": 0.09292944613844156,
      "rewards/rejected": -0.0772908657156222,
      "step": 30
    },
    {
      "entropy": 0.9579024501144886,
      "epoch": 0.5254515599343186,
      "grad_norm": 7.84375,
      "learning_rate": 2.606103007990371e-07,
      "logits/chosen": -1.822963703035332,
      "logits/rejected": -1.8331565552576632,
      "logps/chosen": -365.0903636932373,
      "logps/rejected": -355.14615325927736,
      "loss": 0.6380878925323487,
      "mean_token_accuracy": 0.749091599136591,
      "num_tokens": 5164485.0,
      "rewards/accuracies": 0.7625,
      "rewards/chosen": 0.013401065368088893,
      "rewards/margins": 0.12074792645871639,
      "rewards/rejected": -0.10734686049545417,
      "step": 40
    },
    {
      "entropy": 0.9612626571208238,
      "epoch": 0.6568144499178982,
      "grad_norm": 6.09375,
      "learning_rate": 1.5678588055492286e-07,
      "logits/chosen": -1.8040991478766433,
      "logits/rejected": -1.833159321438822,
      "logps/chosen": -402.17723083496094,
      "logps/rejected": -356.4602531433105,
      "loss": 0.6306396484375,
      "mean_token_accuracy": 0.7441606149077415,
      "num_tokens": 6453542.0,
      "rewards/accuracies": 0.78125,
      "rewards/chosen": 0.014387460224679672,
      "rewards/margins": 0.13786569883814082,
      "rewards/rejected": -0.12347823897434865,
      "step": 50
    },
    {
      "entropy": 0.9673831064254046,
      "epoch": 0.7881773399014779,
      "grad_norm": 6.8125,
      "learning_rate": 6.951097651136889e-08,
      "logits/chosen": -1.8032587367384516,
      "logits/rejected": -1.8365809013216652,
      "logps/chosen": -381.1871561050415,
      "logps/rejected": -348.20880632400514,
      "loss": 0.6237552642822266,
      "mean_token_accuracy": 0.7467762563377619,
      "num_tokens": 7742778.0,
      "rewards/accuracies": 0.79375,
      "rewards/chosen": 0.02942158783553168,
      "rewards/margins": 0.15444288045400753,
      "rewards/rejected": -0.12502129255008185,
      "step": 60
    },
    {
      "entropy": 0.9571243241429329,
      "epoch": 0.9195402298850575,
      "grad_norm": 6.78125,
      "learning_rate": 1.4280638634728948e-08,
      "logits/chosen": -1.8158756551653077,
      "logits/rejected": -1.8351591651305355,
      "logps/chosen": -369.4532477378845,
      "logps/rejected": -360.18324394226073,
      "loss": 0.6174001693725586,
      "mean_token_accuracy": 0.7488324739038944,
      "num_tokens": 9038143.0,
      "rewards/accuracies": 0.8125,
      "rewards/chosen": 0.02844569750013761,
      "rewards/margins": 0.1687723191542318,
      "rewards/rejected": -0.14032662139070454,
      "step": 70
    },
    {
      "epoch": 1.0,
      "eval_entropy": 0.9807351431617998,
      "eval_logits/chosen": -1.8414135470736719,
      "eval_logits/rejected": -1.8220581632231228,
      "eval_logps/chosen": -395.92160922533844,
      "eval_logps/rejected": -371.32895587241813,
      "eval_loss": 0.6231179237365723,
      "eval_mean_token_accuracy": 0.7525672818699928,
      "eval_num_tokens": 9825739.0,
      "eval_rewards/accuracies": 0.7602739726027398,
      "eval_rewards/chosen": 0.013094972891650125,
      "eval_rewards/margins": 0.15859890370693516,
      "eval_rewards/rejected": -0.1455039322377846,
      "eval_runtime": 256.8556,
      "eval_samples_per_second": 1.137,
      "eval_steps_per_second": 0.568,
      "step": 77
    }
  ],
  "logging_steps": 10,
  "max_steps": 77,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 200,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 4.73228507514667e+17,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}