sp-3b-100m/checkpoint-100/trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.5262021973052692,
  "global_step": 100,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [],
  "max_steps": 190,
  "num_train_epochs": 1,
  "total_flos": 8.581652284440576e+17,
  "trial_name": null,
  "trial_params": null
}