{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.9411362080766599,
  "global_step": 5500,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.09,
      "learning_rate": 3.65845311430527e-06,
      "loss": 3.1269,
      "step": 500
    },
    {
      "epoch": 0.17,
      "learning_rate": 3.3162217659137577e-06,
      "loss": 2.719,
      "step": 1000
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.974674880219028e-06,
      "loss": 2.4981,
      "step": 1500
    },
    {
      "epoch": 0.34,
      "learning_rate": 2.6331279945242985e-06,
      "loss": 2.4003,
      "step": 2000
    },
    {
      "epoch": 0.43,
      "learning_rate": 2.2908966461327854e-06,
      "loss": 2.3259,
      "step": 2500
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.948665297741273e-06,
      "loss": 2.2808,
      "step": 3000
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.6064339493497602e-06,
      "loss": 2.2289,
      "step": 3500
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.2642026009582478e-06,
      "loss": 2.1621,
      "step": 4000
    },
    {
      "epoch": 0.77,
      "learning_rate": 9.219712525667351e-07,
      "loss": 2.1678,
      "step": 4500
    },
    {
      "epoch": 0.86,
      "learning_rate": 5.797399041752225e-07,
      "loss": 2.1268,
      "step": 5000
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.3750855578370978e-07,
      "loss": 2.1226,
      "step": 5500
    }
  ],
  "max_steps": 5844,
  "num_train_epochs": 1,
  "total_flos": 7.666297742229504e+16,
  "trial_name": null,
  "trial_params": null
}