chunked/nltk_chunking/_training/chunk_56.txt

Although you can write your own
tf.data pipeline if you want, we have two convenience methods for doing this:

[~TFPreTrainedModel.prepare_tf_dataset]: This is the method we recommend in most cases.