checkpoint-100/trainer_state.json

{
  "best_metric": 0.47333332896232605,
  "best_model_checkpoint": "./zephyr/09-04-24-Weni-WeniGPT-Agents-Zephyr-1.0.15-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": 0.0001785714285714286,
      "loss": 0.4711,
      "step": 20,
      "train/kl": 5.843189239501953,
      "train/logps/chosen": -1172.703515625,
      "train/logps/rejected": -1156.49736328125,
      "train/rewards/chosen": -88.494970703125,
      "train/rewards/margins": -1.6718872070312614,
      "train/rewards/rejected": -86.82308349609374
    },
    {
      "epoch": 0.27,
      "grad_norm": 0.0,
      "learning_rate": 0.00015000000000000001,
      "loss": 0.4437,
      "step": 40,
      "train/kl": 0.0,
      "train/logps/chosen": -2424.414392605634,
      "train/logps/rejected": -2336.605688202247,
      "train/rewards/chosen": -215.21177651848592,
      "train/rewards/margins": -9.581421883654457,
      "train/rewards/rejected": -205.63035463483146
    },
    {
      "epoch": 0.34,
      "eval/kl": 0.0,
      "eval/logps/chosen": -2342.893926056338,
      "eval/logps/rejected": -2118.181566455696,
      "eval/rewards/chosen": -205.92391065140845,
      "eval/rewards/margins": -20.54295238400337,
      "eval/rewards/rejected": -185.38095826740508,
      "eval_loss": 0.47333332896232605,
      "eval_runtime": 140.8471,
      "eval_samples_per_second": 2.13,
      "eval_steps_per_second": 0.532,
      "step": 50
    },
    {
      "epoch": 0.41,
      "grad_norm": 0.0,
      "learning_rate": 0.00012142857142857143,
      "loss": 0.4594,
      "step": 60,
      "train/kl": 0.0,
      "train/logps/chosen": -2388.497661564626,
      "train/logps/rejected": -2377.544617052023,
      "train/rewards/chosen": -211.77136479591837,
      "train/rewards/margins": -3.213425850831669,
      "train/rewards/rejected": -208.5579389450867
    },
    {
      "epoch": 0.55,
      "grad_norm": 0.0,
      "learning_rate": 9.285714285714286e-05,
      "loss": 0.4656,
      "step": 80,
      "train/kl": 0.0,
      "train/logps/chosen": -2349.194211409396,
      "train/logps/rejected": -2324.7878289473683,
      "train/rewards/chosen": -207.3123295931208,
      "train/rewards/margins": -3.7029089059863054,
      "train/rewards/rejected": -203.6094206871345
    },
    {
      "epoch": 0.68,
      "grad_norm": 0.0,
      "learning_rate": 6.428571428571429e-05,
      "loss": 0.4469,
      "step": 100,
      "train/kl": 0.0,
      "train/logps/chosen": -2737.1844405594406,
      "train/logps/rejected": -2257.276836158192,
      "train/rewards/chosen": -243.65840799825176,
      "train/rewards/margins": -45.40216241209356,
      "train/rewards/rejected": -198.2562455861582
    },
    {
      "epoch": 0.68,
      "eval/kl": 0.0,
      "eval/logps/chosen": -2343.730193661972,
      "eval/logps/rejected": -2118.9036787974683,
      "eval/rewards/chosen": -206.00756492077466,
      "eval/rewards/margins": -20.55440283612276,
      "eval/rewards/rejected": -185.4531620846519,
      "eval_loss": 0.47333332896232605,
      "eval_runtime": 140.7532,
      "eval_samples_per_second": 2.131,
      "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
}