checkpoint-100/trainer_state.json

{
  "best_metric": 0.6417509913444519,
  "best_model_checkpoint": "./Zephyr/14-03-24-Weni-WeniGPT-2.4.1-Zephyr-7B-5-epochs-LLM_Base_2.0.3_DPO_WeniGPT DPO training-2_max_steps-445_batch_32_2024-03-14_ppid_9/checkpoint-100",
  "epoch": 1.1173184357541899,
  "eval_steps": 100,
  "global_step": 100,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.22,
      "grad_norm": 1.764265537261963,
      "learning_rate": 2.2222222222222223e-05,
      "logits/chosen": -2.2127978801727295,
      "logits/rejected": -2.2138168811798096,
      "logps/chosen": -6.020094871520996,
      "logps/rejected": -10.437583923339844,
      "loss": 0.6881,
      "rewards/accuracies": 0.05781250074505806,
      "rewards/chosen": 0.008710675872862339,
      "rewards/margins": 0.011647692881524563,
      "rewards/rejected": -0.002937017474323511,
      "step": 20
    },
    {
      "epoch": 0.45,
      "grad_norm": 0.008939280174672604,
      "learning_rate": 4.4444444444444447e-05,
      "logits/chosen": -2.209024429321289,
      "logits/rejected": -2.209834575653076,
      "logps/chosen": -4.6792497634887695,
      "logps/rejected": -9.106904983520508,
      "loss": 0.6701,
      "rewards/accuracies": 0.04531250149011612,
      "rewards/chosen": 0.09174498170614243,
      "rewards/margins": 0.13157977163791656,
      "rewards/rejected": -0.039834775030612946,
      "step": 40
    },
    {
      "epoch": 0.67,
      "grad_norm": 1.14188189570541e-07,
      "learning_rate": 4.8125000000000004e-05,
      "logits/chosen": -2.140080213546753,
      "logits/rejected": -2.140465259552002,
      "logps/chosen": -5.5485029220581055,
      "logps/rejected": -12.926416397094727,
      "loss": 0.6585,
      "rewards/accuracies": 0.05156249925494194,
      "rewards/chosen": -0.03698154538869858,
      "rewards/margins": 0.39537638425827026,
      "rewards/rejected": -0.43235793709754944,
      "step": 60
    },
    {
      "epoch": 0.89,
      "grad_norm": 16.8149471282959,
      "learning_rate": 4.575e-05,
      "logits/chosen": -2.026047945022583,
      "logits/rejected": -2.025956392288208,
      "logps/chosen": -10.022688865661621,
      "logps/rejected": -24.660259246826172,
      "loss": 0.6669,
      "rewards/accuracies": 0.05781250074505806,
      "rewards/chosen": -0.3339082598686218,
      "rewards/margins": 1.0433582067489624,
      "rewards/rejected": -1.377266526222229,
      "step": 80
    },
    {
      "epoch": 1.12,
      "grad_norm": 0.0,
      "learning_rate": 4.325e-05,
      "logits/chosen": -1.9624429941177368,
      "logits/rejected": -1.9612023830413818,
      "logps/chosen": -10.355883598327637,
      "logps/rejected": -26.145648956298828,
      "loss": 0.6643,
      "rewards/accuracies": 0.05312500149011612,
      "rewards/chosen": -0.4869672358036041,
      "rewards/margins": 1.212210774421692,
      "rewards/rejected": -1.6991779804229736,
      "step": 100
    },
    {
      "epoch": 1.12,
      "eval_logits/chosen": -1.8881776332855225,
      "eval_logits/rejected": -1.8856115341186523,
      "eval_logps/chosen": -17.27315902709961,
      "eval_logps/rejected": -45.2144889831543,
      "eval_loss": 0.6417509913444519,
      "eval_rewards/accuracies": 0.07604166865348816,
      "eval_rewards/chosen": -0.8916252851486206,
      "eval_rewards/margins": 2.1516268253326416,
      "eval_rewards/rejected": -3.0432522296905518,
      "eval_runtime": 183.8919,
      "eval_samples_per_second": 1.729,
      "eval_steps_per_second": 0.218,
      "step": 100
    }
  ],
  "logging_steps": 20,
  "max_steps": 445,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 100,
  "total_flos": 0.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}