checkpoint-100/trainer_state.json

{
  "best_metric": 0.47333332896232605,
  "best_model_checkpoint": "./zephyr/05-04-24-Weni-WeniGPT-Agents-Zephyr-1.0.1-KTO_testing kto dataset during training-3_max_steps-145_batch_16_2024-04-05_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": 4.223984241485596,
      "kl": 50.472633361816406,
      "learning_rate": 0.00019,
      "logps/chosen": -516.6061401367188,
      "logps/rejected": -730.8901977539062,
      "loss": 0.4891,
      "rewards/chosen": -2.2866313457489014,
      "rewards/margins": 9.197293281555176,
      "rewards/rejected": -11.027228355407715,
      "step": 20
    },
    {
      "epoch": 0.27,
      "grad_norm": 4.9320197105407715,
      "kl": 227.55972290039062,
      "learning_rate": 0.00016142857142857145,
      "logps/chosen": -400.697998046875,
      "logps/rejected": -375.78631591796875,
      "loss": 0.5465,
      "rewards/chosen": 8.534103393554688,
      "rewards/margins": -13.850674629211426,
      "rewards/rejected": 23.049671173095703,
      "step": 40
    },
    {
      "epoch": 0.34,
      "eval_kl": 45.850624084472656,
      "eval_logps/chosen": -710.6041870117188,
      "eval_logps/rejected": -693.1708374023438,
      "eval_loss": 0.5112424492835999,
      "eval_rewards/chosen": -20.341365814208984,
      "eval_rewards/margins": -17.1827449798584,
      "eval_rewards/rejected": -4.146603107452393,
      "eval_runtime": 138.4214,
      "eval_samples_per_second": 2.167,
      "eval_steps_per_second": 0.542,
      "step": 50
    },
    {
      "epoch": 0.41,
      "grad_norm": 0.0030494395177811384,
      "kl": 210.3240966796875,
      "learning_rate": 0.00013285714285714287,
      "logps/chosen": -414.4272155761719,
      "logps/rejected": -540.6943969726562,
      "loss": 0.4697,
      "rewards/chosen": 17.047597885131836,
      "rewards/margins": 2.3560233116149902,
      "rewards/rejected": 15.122901916503906,
      "step": 60
    },
    {
      "epoch": 0.55,
      "grad_norm": 52.698368072509766,
      "kl": 257.76116943359375,
      "learning_rate": 0.0001042857142857143,
      "logps/chosen": -320.7880859375,
      "logps/rejected": -430.06829833984375,
      "loss": 0.5455,
      "rewards/chosen": 19.29902458190918,
      "rewards/margins": -8.448366165161133,
      "rewards/rejected": 28.257070541381836,
      "step": 80
    },
    {
      "epoch": 0.68,
      "grad_norm": 0.0,
      "kl": 40.73902130126953,
      "learning_rate": 7.571428571428571e-05,
      "logps/chosen": -1188.35693359375,
      "logps/rejected": -1296.0433349609375,
      "loss": 0.4745,
      "rewards/chosen": -56.30565643310547,
      "rewards/margins": 8.954660415649414,
      "rewards/rejected": -68.27855682373047,
      "step": 100
    },
    {
      "epoch": 0.68,
      "eval_kl": 0.0,
      "eval_logps/chosen": -2394.681884765625,
      "eval_logps/rejected": -2418.263916015625,
      "eval_loss": 0.47333332896232605,
      "eval_rewards/chosen": -188.7490997314453,
      "eval_rewards/margins": -12.401326179504395,
      "eval_rewards/rejected": -176.65597534179688,
      "eval_runtime": 138.4851,
      "eval_samples_per_second": 2.166,
      "eval_steps_per_second": 0.542,
      "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
}