{
  "best_global_step": 146,
  "best_metric": 9972.7021484375,
  "best_model_checkpoint": "/NHNHOME/WORKSPACE/26msit001_A/mnpo/revision_qwen3_8b/full_iter1/train/inpo_avg_eta0p005_s42_nhn-b200-q3-seq-07101430-inpo/checkpoint-146",
  "epoch": 1.0,
  "eval_steps": 500,
  "global_step": 146,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.06864006864006864,
      "grad_norm": 20387.03686881509,
      "learning_rate": 4.990077890125363e-07,
      "logits/chosen": -0.796875,
      "logits/rejected": -0.8046875,
      "logps/chosen": -16.0,
      "logps/rejected": -15.9375,
      "loss": 9979.939453125,
      "rewards/accuracies": 0.518750011920929,
      "rewards/chosen": -153.50888061523438,
      "rewards/margins": 1.0440444946289062,
      "rewards/rejected": -154.55291748046875,
      "step": 10
    },
    {
      "epoch": 0.13728013728013727,
      "grad_norm": 16463.56555342243,
      "learning_rate": 4.879356673988089e-07,
      "logits/chosen": -0.87109375,
      "logits/rejected": -0.84765625,
      "logps/chosen": -15.8125,
      "logps/rejected": -15.875,
      "loss": 9975.23828125,
      "rewards/accuracies": 0.5249999761581421,
      "rewards/chosen": -153.51373291015625,
      "rewards/margins": 1.252532958984375,
      "rewards/rejected": -154.76626586914062,
      "step": 20
    },
    {
      "epoch": 0.2059202059202059,
      "grad_norm": 28044.862093220032,
      "learning_rate": 4.6510039481503485e-07,
      "logits/chosen": -0.74609375,
      "logits/rejected": -0.7421875,
      "logps/chosen": -16.0,
      "logps/rejected": -16.0,
      "loss": 9996.552734375,
      "rewards/accuracies": 0.48124998807907104,
      "rewards/chosen": -155.00054931640625,
      "rewards/margins": 0.214385986328125,
      "rewards/rejected": -155.21493530273438,
      "step": 30
    },
    {
      "epoch": 0.27456027456027454,
      "grad_norm": 19615.09853150976,
      "learning_rate": 4.3163090985954074e-07,
      "logits/chosen": -0.734375,
      "logits/rejected": -0.74609375,
      "logps/chosen": -16.0,
      "logps/rejected": -16.0,
      "loss": 9994.1796875,
      "rewards/accuracies": 0.53125,
      "rewards/chosen": -155.23875427246094,
      "rewards/margins": 0.32525634765625,
      "rewards/rejected": -155.56399536132812,
      "step": 40
    },
    {
      "epoch": 0.3432003432003432,
      "grad_norm": 18075.561008642915,
      "learning_rate": 3.891818892301304e-07,
      "logits/chosen": -0.71484375,
      "logits/rejected": -0.71875,
      "logps/chosen": -16.125,
      "logps/rejected": -16.125,
      "loss": 9973.58203125,
      "rewards/accuracies": 0.518750011920929,
      "rewards/chosen": -154.60614013671875,
      "rewards/margins": 1.3658523559570312,
      "rewards/rejected": -155.9720001220703,
      "step": 50
    },
    {
      "epoch": 0.4118404118404118,
      "grad_norm": 21829.673229552107,
      "learning_rate": 3.3985194320937815e-07,
      "logits/chosen": -0.6953125,
      "logits/rejected": -0.6875,
      "logps/chosen": -16.125,
      "logps/rejected": -16.125,
      "loss": 9989.525390625,
      "rewards/accuracies": 0.550000011920929,
      "rewards/chosen": -156.18487548828125,
      "rewards/margins": 0.5427627563476562,
      "rewards/rejected": -156.72763061523438,
      "step": 60
    },
    {
      "epoch": 0.4804804804804805,
      "grad_norm": 19354.982747194757,
      "learning_rate": 2.8607986379820664e-07,
      "logits/chosen": -0.66015625,
      "logits/rejected": -0.6484375,
      "logps/chosen": -16.0,
      "logps/rejected": -16.125,
      "loss": 9972.458984375,
      "rewards/accuracies": 0.581250011920929,
      "rewards/chosen": -155.47698974609375,
      "rewards/margins": 1.410614013671875,
      "rewards/rejected": -156.88760375976562,
      "step": 70
    },
    {
      "epoch": 0.5491205491205491,
      "grad_norm": 23696.769639901566,
      "learning_rate": 2.3052405482064919e-07,
      "logits/chosen": -0.65625,
      "logits/rejected": -0.67578125,
      "logps/chosen": -16.125,
      "logps/rejected": -16.0,
      "loss": 9998.822265625,
      "rewards/accuracies": 0.4749999940395355,
      "rewards/chosen": -155.87271118164062,
      "rewards/margins": 0.087371826171875,
      "rewards/rejected": -155.9600830078125,
      "step": 80
    },
    {
      "epoch": 0.6177606177606177,
      "grad_norm": 40006.950593012094,
      "learning_rate": 1.7593110477859152e-07,
      "logits/chosen": -0.67578125,
      "logits/rejected": -0.671875,
      "logps/chosen": -16.0,
      "logps/rejected": -16.0,
      "loss": 9985.57421875,
      "rewards/accuracies": 0.543749988079071,
      "rewards/chosen": -154.62530517578125,
      "rewards/margins": 0.742950439453125,
      "rewards/rejected": -155.36825561523438,
      "step": 90
    },
    {
      "epoch": 0.6864006864006864,
      "grad_norm": 18244.96636423906,
      "learning_rate": 1.2500000000000005e-07,
      "logits/chosen": -0.62890625,
      "logits/rejected": -0.6171875,
      "logps/chosen": -16.0,
      "logps/rejected": -16.0,
      "loss": 9980.126953125,
      "rewards/accuracies": 0.5874999761581421,
      "rewards/chosen": -154.44403076171875,
      "rewards/margins": 1.02630615234375,
      "rewards/rejected": -155.47032165527344,
      "step": 100
    },
    {
      "epoch": 0.7550407550407551,
      "grad_norm": 27502.551311382453,
      "learning_rate": 8.024869116091879e-08,
      "logits/chosen": -0.6484375,
      "logits/rejected": -0.62890625,
      "logps/chosen": -16.0,
      "logps/rejected": -16.125,
      "loss": 9976.1513671875,
      "rewards/accuracies": 0.5687500238418579,
      "rewards/chosen": -155.1019287109375,
      "rewards/margins": 1.234588623046875,
      "rewards/rejected": -156.33651733398438,
      "step": 110
    },
    {
      "epoch": 0.8236808236808236,
      "grad_norm": 19647.533994105936,
      "learning_rate": 4.388960991455998e-08,
      "logits/chosen": -0.58984375,
      "logits/rejected": -0.59375,
      "logps/chosen": -16.125,
      "logps/rejected": -16.125,
      "loss": 9963.115234375,
      "rewards/accuracies": 0.550000011920929,
      "rewards/chosen": -154.8621826171875,
      "rewards/margins": 1.900604248046875,
      "rewards/rejected": -156.76278686523438,
      "step": 120
    },
    {
      "epoch": 0.8923208923208923,
      "grad_norm": 20937.34630051602,
      "learning_rate": 1.7720289882128092e-08,
      "logits/chosen": -0.62109375,
      "logits/rejected": -0.609375,
      "logps/chosen": -16.0,
      "logps/rejected": -16.0,
      "loss": 9983.845703125,
      "rewards/accuracies": 0.53125,
      "rewards/chosen": -155.11953735351562,
      "rewards/margins": 0.8319244384765625,
      "rewards/rejected": -155.9514617919922,
      "step": 130
    },
    {
      "epoch": 0.960960960960961,
      "grad_norm": 20078.25155988699,
      "learning_rate": 3.0344995250326245e-09,
      "logits/chosen": -0.578125,
      "logits/rejected": -0.58984375,
      "logps/chosen": -16.0,
      "logps/rejected": -16.0,
      "loss": 9975.1064453125,
      "rewards/accuracies": 0.5562499761581421,
      "rewards/chosen": -155.063720703125,
      "rewards/margins": 1.2694091796875,
      "rewards/rejected": -156.3331298828125,
      "step": 140
    },
    {
      "epoch": 1.0,
      "eval_logits/chosen": -0.63671875,
      "eval_logits/rejected": -0.60546875,
      "eval_logps/chosen": -15.9375,
      "eval_logps/rejected": -16.0,
      "eval_loss": 9972.7021484375,
      "eval_rewards/accuracies": 0.5526315569877625,
      "eval_rewards/chosen": -155.34686279296875,
      "eval_rewards/margins": 1.3776253461837769,
      "eval_rewards/rejected": -156.7244873046875,
      "eval_runtime": 68.5896,
      "eval_samples_per_second": 8.806,
      "eval_steps_per_second": 1.108,
      "step": 146
    },
    {
      "epoch": 1.0,
      "step": 146,
      "total_flos": 0.0,
      "train_loss": 9980.011584974314,
      "train_runtime": 3282.8375,
      "train_samples_per_second": 5.679,
      "train_steps_per_second": 0.044
    }
  ],
  "logging_steps": 10,
  "max_steps": 146,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 1000,
  "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": 1,
  "trial_name": null,
  "trial_params": null
}