checkpoint-100/trainer_state.json

{
  "best_metric": 0.47333332896232605,
  "best_model_checkpoint": "./zephyr/08-04-24-Weni-WeniGPT-Agents-Zephyr-1.0.16-KTO_Hyperparameter search, altering lora params for KTO task.-2_max_steps-145_batch_16_2024-04-08_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": 0.4847,
      "step": 20,
      "train/kl": 0.0,
      "train/logps/chosen": -1869.3137335526317,
      "train/logps/rejected": -1463.0350632440477,
      "train/rewards/chosen": -157.77652138157896,
      "train/rewards/margins": -39.35788996416825,
      "train/rewards/rejected": -118.41863141741071
    },
    {
      "epoch": 0.27,
      "grad_norm": 0.0,
      "learning_rate": 0.00015142857142857143,
      "loss": 0.4938,
      "step": 40,
      "train/kl": 0.0,
      "train/logps/chosen": -2339.678006329114,
      "train/logps/rejected": -2094.7118055555557,
      "train/rewards/chosen": -204.7826839398734,
      "train/rewards/margins": -23.895109190645,
      "train/rewards/rejected": -180.8875747492284
    },
    {
      "epoch": 0.34,
      "eval/kl": 0.0,
      "eval/logps/chosen": -2160.559198943662,
      "eval/logps/rejected": -1937.6455696202531,
      "eval/rewards/chosen": -187.69000330105635,
      "eval/rewards/margins": -20.362446932385467,
      "eval/rewards/rejected": -167.32755636867088,
      "eval_loss": 0.47333332896232605,
      "eval_runtime": 141.6142,
      "eval_samples_per_second": 2.118,
      "eval_steps_per_second": 0.53,
      "step": 50
    },
    {
      "epoch": 0.41,
      "grad_norm": 0.0,
      "learning_rate": 0.00012285714285714287,
      "loss": 0.5,
      "step": 60,
      "train/kl": 0.0,
      "train/logps/chosen": -2174.43828125,
      "train/logps/rejected": -2130.2125,
      "train/rewards/chosen": -189.318359375,
      "train/rewards/margins": -4.711669921875,
      "train/rewards/rejected": -184.606689453125
    },
    {
      "epoch": 0.55,
      "grad_norm": 0.0,
      "learning_rate": 9.428571428571429e-05,
      "loss": 0.4562,
      "step": 80,
      "train/kl": 0.0,
      "train/logps/chosen": -2384.5894691780823,
      "train/logps/rejected": -1893.6566091954023,
      "train/rewards/chosen": -209.14641748715752,
      "train/rewards/margins": -46.95111621201383,
      "train/rewards/rejected": -162.1953012751437
    },
    {
      "epoch": 0.68,
      "grad_norm": 0.0,
      "learning_rate": 6.571428571428571e-05,
      "loss": 0.4562,
      "step": 100,
      "train/kl": 0.0,
      "train/logps/chosen": -2356.8424657534247,
      "train/logps/rejected": -2293.712643678161,
      "train/rewards/chosen": -207.19579409246575,
      "train/rewards/margins": -8.242444595339322,
      "train/rewards/rejected": -198.95334949712642
    },
    {
      "epoch": 0.68,
      "eval/kl": 0.0,
      "eval/logps/chosen": -2160.684639084507,
      "eval/logps/rejected": -1938.0195806962026,
      "eval/rewards/chosen": -187.70257482394365,
      "eval/rewards/margins": -20.337624764608194,
      "eval/rewards/rejected": -167.36495005933546,
      "eval_loss": 0.47333332896232605,
      "eval_runtime": 141.5959,
      "eval_samples_per_second": 2.119,
      "eval_steps_per_second": 0.53,
      "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
}