checkpoint-100/trainer_state.json

{
  "best_metric": 0.5,
  "best_model_checkpoint": "./zephyr/10-04-24-Weni-WeniGPT-Agents-Zephyr-1.0.21-KTO_Experiment with a new tokenizer configuration for chat template of zephyr-2_max_steps-147_batch_16_2024-04-10_ppid_9/checkpoint-100",
  "epoch": 0.6600660066006601,
  "eval_steps": 50,
  "global_step": 100,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.13,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 0.0001859154929577465,
      "logps/chosen": NaN,
      "logps/rejected": NaN,
      "loss": 0.6544,
      "rewards/chosen": NaN,
      "rewards/margins": NaN,
      "rewards/rejected": NaN,
      "step": 20
    },
    {
      "epoch": 0.26,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 0.00015774647887323943,
      "logps/chosen": NaN,
      "logps/rejected": NaN,
      "loss": 0.6951,
      "rewards/chosen": NaN,
      "rewards/margins": NaN,
      "rewards/rejected": NaN,
      "step": 40
    },
    {
      "epoch": 0.33,
      "eval_kl": 0.0,
      "eval_logps/chosen": -2315.623779296875,
      "eval_logps/rejected": -1935.1795654296875,
      "eval_loss": 0.5,
      "eval_rewards/chosen": -203.61370849609375,
      "eval_rewards/margins": -33.913272857666016,
      "eval_rewards/rejected": -169.70042419433594,
      "eval_runtime": 171.3306,
      "eval_samples_per_second": 2.043,
      "eval_steps_per_second": 0.514,
      "step": 50
    },
    {
      "epoch": 0.4,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 0.0001295774647887324,
      "logps/chosen": NaN,
      "logps/rejected": NaN,
      "loss": 0.7834,
      "rewards/chosen": NaN,
      "rewards/margins": NaN,
      "rewards/rejected": NaN,
      "step": 60
    },
    {
      "epoch": 0.53,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 0.00010140845070422535,
      "logps/chosen": NaN,
      "logps/rejected": NaN,
      "loss": 0.6729,
      "rewards/chosen": NaN,
      "rewards/margins": NaN,
      "rewards/rejected": NaN,
      "step": 80
    },
    {
      "epoch": 0.66,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 7.323943661971832e-05,
      "logps/chosen": NaN,
      "logps/rejected": NaN,
      "loss": 0.6788,
      "rewards/chosen": NaN,
      "rewards/margins": NaN,
      "rewards/rejected": NaN,
      "step": 100
    },
    {
      "epoch": 0.66,
      "eval_kl": 0.0,
      "eval_logps/chosen": -2304.287109375,
      "eval_logps/rejected": -1926.1640625,
      "eval_loss": 0.5,
      "eval_rewards/chosen": -202.4800262451172,
      "eval_rewards/margins": -33.68117904663086,
      "eval_rewards/rejected": -168.79885864257812,
      "eval_runtime": 171.6567,
      "eval_samples_per_second": 2.039,
      "eval_steps_per_second": 0.513,
      "step": 100
    }
  ],
  "logging_steps": 20,
  "max_steps": 147,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 100,
  "total_flos": 0.0,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}