trainer_state.json

{
  "best_metric": 0.38227128982543945,
  "best_model_checkpoint": "model_result/02_02_SoLAR_10.7B_DPO/checkpoint-2757",
  "epoch": 2.999592003263974,
  "eval_steps": 919,
  "global_step": 2757,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.9998640010879913,
      "grad_norm": 29.002643585205078,
      "learning_rate": 4.4449818621523576e-07,
      "logits/chosen": -1.361188530921936,
      "logits/rejected": -1.3760977983474731,
      "logps/chosen": -369.7716064453125,
      "logps/rejected": -341.9248352050781,
      "loss": 0.9991,
      "rewards/accuracies": 0.6764485836029053,
      "rewards/chosen": 20.894943237304688,
      "rewards/margins": 2.2651174068450928,
      "rewards/rejected": 18.629825592041016,
      "step": 919
    },
    {
      "epoch": 0.9998640010879913,
      "eval_logits/chosen": -1.3181335926055908,
      "eval_logits/rejected": -1.346123456954956,
      "eval_logps/chosen": -527.0983276367188,
      "eval_logps/rejected": -507.80621337890625,
      "eval_loss": 0.49919795989990234,
      "eval_rewards/accuracies": 0.8106383085250854,
      "eval_rewards/chosen": 5.702141761779785,
      "eval_rewards/margins": 3.5380544662475586,
      "eval_rewards/rejected": 2.1640872955322266,
      "eval_runtime": 456.9799,
      "eval_samples_per_second": 2.053,
      "eval_steps_per_second": 0.514,
      "step": 919
    },
    {
      "epoch": 1.9997280021759827,
      "grad_norm": 29.223295211791992,
      "learning_rate": 3.3337363966142684e-07,
      "logits/chosen": -1.3160151243209839,
      "logits/rejected": -1.3324289321899414,
      "logps/chosen": -529.5614624023438,
      "logps/rejected": -521.8484497070312,
      "loss": 0.3824,
      "rewards/accuracies": 0.8457222580909729,
      "rewards/chosen": 4.916207790374756,
      "rewards/margins": 4.2801947593688965,
      "rewards/rejected": 0.6360131502151489,
      "step": 1838
    },
    {
      "epoch": 1.9997280021759827,
      "eval_logits/chosen": -1.3160426616668701,
      "eval_logits/rejected": -1.3441028594970703,
      "eval_logps/chosen": -559.829345703125,
      "eval_logps/rejected": -548.035400390625,
      "eval_loss": 0.39502739906311035,
      "eval_rewards/accuracies": 0.8446808457374573,
      "eval_rewards/chosen": 2.4290404319763184,
      "eval_rewards/margins": 4.287874221801758,
      "eval_rewards/rejected": -1.858833909034729,
      "eval_runtime": 456.8696,
      "eval_samples_per_second": 2.053,
      "eval_steps_per_second": 0.514,
      "step": 1838
    },
    {
      "epoch": 2.999592003263974,
      "grad_norm": 18.053089141845703,
      "learning_rate": 2.2224909310761788e-07,
      "logits/chosen": -1.3009860515594482,
      "logits/rejected": -1.3179205656051636,
      "logps/chosen": -557.4498291015625,
      "logps/rejected": -556.6650390625,
      "loss": 0.2672,
      "rewards/accuracies": 0.8876156210899353,
      "rewards/chosen": 2.127683401107788,
      "rewards/margins": 4.969925403594971,
      "rewards/rejected": -2.8422415256500244,
      "step": 2757
    },
    {
      "epoch": 2.999592003263974,
      "eval_logits/chosen": -1.2920633554458618,
      "eval_logits/rejected": -1.320629596710205,
      "eval_logps/chosen": -571.9027709960938,
      "eval_logps/rejected": -563.9360961914062,
      "eval_loss": 0.38227128982543945,
      "eval_rewards/accuracies": 0.8542553186416626,
      "eval_rewards/chosen": 1.2216957807540894,
      "eval_rewards/margins": 4.6706061363220215,
      "eval_rewards/rejected": -3.4489102363586426,
      "eval_runtime": 456.9375,
      "eval_samples_per_second": 2.053,
      "eval_steps_per_second": 0.514,
      "step": 2757
    }
  ],
  "logging_steps": 919,
  "max_steps": 4595,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 919,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 0.0,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}