Add use_auth to load_datasets for private datasets to PT and TF examples (#16521)

* fix formatting and remove use_auth

* Add use_auth_token to Flax examples

examples/pytorch/image-pretraining/run_mae.py

@@ -207,6 +207,7 @@ def main():
         data_args.dataset_config_name,
         data_files=data_args.data_files,
         cache_dir=model_args.cache_dir,
+        use_auth_token=True if model_args.use_auth_token else None,
     )
 
     # If we don't have a validation split, split off a percentage of train as validation.