checkpoint-100/trainer_state.json

{
  "best_metric": 0.47333332896232605,
  "best_model_checkpoint": "./zephyr/09-04-24-Weni-WeniGPT-Agents-Zephyr-1.0.16-KTO_Hyperparameter search, altering lora params for KTO task.-2_max_steps-145_batch_16_2024-04-09_ppid_9/checkpoint-100",
  "epoch": 0.684931506849315,
  "eval_steps": 50,
  "global_step": 100,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.14,
      "grad_norm": 0.0,
      "learning_rate": 0.00018,
      "loss": 0.4713,
      "step": 20,
      "train/kl": 0.0,
      "train/logps/chosen": -1423.408907312925,
      "train/logps/rejected": -1301.3419436416184,
      "train/rewards/chosen": -113.72140731292517,
      "train/rewards/margins": -11.091552724774871,
      "train/rewards/rejected": -102.6298545881503
    },
    {
      "epoch": 0.27,
      "grad_norm": 0.0,
      "learning_rate": 0.00015142857142857143,
      "loss": 0.45,
      "step": 40,
      "train/kl": 0.0,
      "train/logps/chosen": -2404.9383680555557,
      "train/logps/rejected": -2267.5051491477275,
      "train/rewards/chosen": -211.85392252604166,
      "train/rewards/margins": -15.253491950757564,
      "train/rewards/rejected": -196.6004305752841
    },
    {
      "epoch": 0.34,
      "eval/kl": 0.0,
      "eval/logps/chosen": -2179.756602112676,
      "eval/logps/rejected": -1974.1839398734178,
      "eval/rewards/chosen": -189.60969135123239,
      "eval/rewards/margins": -18.62818423097923,
      "eval/rewards/rejected": -170.98150712025316,
      "eval_loss": 0.47333332896232605,
      "eval_runtime": 140.7497,
      "eval_samples_per_second": 2.131,
      "eval_steps_per_second": 0.533,
      "step": 50
    },
    {
      "epoch": 0.41,
      "grad_norm": 0.0,
      "learning_rate": 0.00012285714285714287,
      "loss": 0.4469,
      "step": 60,
      "train/kl": 0.0,
      "train/logps/chosen": -2463.7814685314684,
      "train/logps/rejected": -2104.4410310734465,
      "train/rewards/chosen": -216.41613854895104,
      "train/rewards/margins": -33.82629479895104,
      "train/rewards/rejected": -182.58984375
    },
    {
      "epoch": 0.55,
      "grad_norm": 0.0,
      "learning_rate": 9.428571428571429e-05,
      "loss": 0.4281,
      "step": 80,
      "train/kl": 0.0,
      "train/logps/chosen": -2216.6425638686133,
      "train/logps/rejected": -2146.411031420765,
      "train/rewards/chosen": -193.77794251824818,
      "train/rewards/margins": -6.814528925351993,
      "train/rewards/rejected": -186.9634135928962
    },
    {
      "epoch": 0.68,
      "grad_norm": 0.0,
      "learning_rate": 6.571428571428571e-05,
      "loss": 0.5031,
      "step": 100,
      "train/kl": 0.0,
      "train/logps/chosen": -2453.4375,
      "train/logps/rejected": -2285.3565251572327,
      "train/rewards/chosen": -214.367794060559,
      "train/rewards/margins": -16.124685275181633,
      "train/rewards/rejected": -198.24310878537736
    },
    {
      "epoch": 0.68,
      "eval/kl": 0.0,
      "eval/logps/chosen": -2181.0286091549297,
      "eval/logps/rejected": -1975.4440268987341,
      "eval/rewards/chosen": -189.73686454665494,
      "eval/rewards/margins": -18.629343779249865,
      "eval/rewards/rejected": -171.10752076740508,
      "eval_loss": 0.47333332896232605,
      "eval_runtime": 140.6959,
      "eval_samples_per_second": 2.132,
      "eval_steps_per_second": 0.533,
      "step": 100
    }
  ],
  "logging_steps": 20,
  "max_steps": 145,
  "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
}