models/scibert/trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.6007802340702213,
  "global_step": 2000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.65,
      "learning_rate": 3.916341569137408e-05,
      "loss": 1.4943,
      "step": 500
    },
    {
      "epoch": 1.3,
      "learning_rate": 2.832683138274816e-05,
      "loss": 1.0861,
      "step": 1000
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.7490247074122236e-05,
      "loss": 0.9443,
      "step": 1500
    },
    {
      "epoch": 2.6,
      "learning_rate": 6.653662765496315e-06,
      "loss": 0.7464,
      "step": 2000
    }
  ],
  "max_steps": 2307,
  "num_train_epochs": 3,
  "total_flos": 1.1652712606267776e+16,
  "trial_name": null,
  "trial_params": null
}