checkpoint-100/trainer_state.json

{
  "best_metric": 0.5443353056907654,
  "best_model_checkpoint": "./zephyr/08-04-24-Weni-WeniGPT-Agents-Zephyr-1.0.9-KTO_Hyperparameter search, altering desired and undesired weights 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": 4.8340067863464355,
      "kl": 27.297277450561523,
      "learning_rate": 0.00018,
      "logps/chosen": -294.6091003417969,
      "logps/rejected": -256.2513732910156,
      "loss": 0.4863,
      "rewards/chosen": 1.882211446762085,
      "rewards/margins": 0.8332540392875671,
      "rewards/rejected": 0.9183141589164734,
      "step": 20
    },
    {
      "epoch": 0.27,
      "grad_norm": 0.0,
      "kl": 40.49668502807617,
      "learning_rate": 0.00015142857142857143,
      "logps/chosen": -539.7681884765625,
      "logps/rejected": -501.2303161621094,
      "loss": 0.4555,
      "rewards/chosen": -24.707237243652344,
      "rewards/margins": -0.5242304801940918,
      "rewards/rejected": -22.829710006713867,
      "step": 40
    },
    {
      "epoch": 0.34,
      "eval_kl": 0.0,
      "eval_logps/chosen": -1786.45947265625,
      "eval_logps/rejected": -1616.440185546875,
      "eval_loss": 0.5443361401557922,
      "eval_rewards/chosen": -148.88607788085938,
      "eval_rewards/margins": -20.042972564697266,
      "eval_rewards/rejected": -135.36317443847656,
      "eval_runtime": 138.4944,
      "eval_samples_per_second": 2.166,
      "eval_steps_per_second": 0.542,
      "step": 50
    },
    {
      "epoch": 0.41,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 0.00012285714285714287,
      "logps/chosen": -1740.7340087890625,
      "logps/rejected": -1820.6246337890625,
      "loss": 0.5427,
      "rewards/chosen": -146.22254943847656,
      "rewards/margins": 5.4553728103637695,
      "rewards/rejected": -151.97308349609375,
      "step": 60
    },
    {
      "epoch": 0.55,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 9.428571428571429e-05,
      "logps/chosen": -1711.9744873046875,
      "logps/rejected": -1807.9703369140625,
      "loss": 0.5427,
      "rewards/chosen": -143.12269592285156,
      "rewards/margins": 8.623757362365723,
      "rewards/rejected": -150.7401885986328,
      "step": 80
    },
    {
      "epoch": 0.68,
      "grad_norm": 0.0,
      "kl": 0.0,
      "learning_rate": 6.571428571428571e-05,
      "logps/chosen": -1719.2364501953125,
      "logps/rejected": -1687.2681884765625,
      "loss": 0.5714,
      "rewards/chosen": -144.7313995361328,
      "rewards/margins": -4.064985752105713,
      "rewards/rejected": -140.0031280517578,
      "step": 100
    },
    {
      "epoch": 0.68,
      "eval_kl": 0.0,
      "eval_logps/chosen": -1605.7635498046875,
      "eval_logps/rejected": -1472.8948974609375,
      "eval_loss": 0.5443353056907654,
      "eval_rewards/chosen": -130.8164825439453,
      "eval_rewards/margins": -16.295190811157227,
      "eval_rewards/rejected": -121.00863647460938,
      "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
}