Tensorflow Serving сохранить модель Deepspell

Question

Я хочу сохранить глубоко корректную модель в Tensorflow Serving, но я получаю ошибку

from deepcorrect import DeepCorrect
model = DeepCorrect('deeppunct_params_en', 'deeppunct_checkpoint_wikipedia')
model.deepcorrect_model[0]
import tempfile
import os
import tensorflow as tf


MODEL_DIR = tempfile.gettempdir()
version = 1
export_path = os.path.join(MODEL_DIR, str(version))
print('export_path = {}\n'.format(export_path))

tf.keras.models.save_model(
    model.deepcorrect_model[0],
    export_path,
    overwrite=True,
    include_optimizer=True,
    save_format=None,
    signatures=None
)

print('\nSaved model:')
!ls -l {export_path}

Это моя ошибка:

export_path = /tmp/1

/usr/local/lib/python3.6/dist-packages/keras/engine/network.py:883: UserWarning: Layer lstm_2 was passed non-serializable keyword arguments: {'initial_state': [<tf.Tensor 'strided_slice:0' shape=(?, 256) dtype=float32>, <tf.Tensor 'strided_slice:0' shape=(?, 256) dtype=float32>]}. They will not be included in the serialized model (and thus will be missing at deserialization time).
  '. They will not be included '
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-6-7837d3d06154> in <module>()
     15     include_optimizer=True,
     16     save_format=None,
---> 17     signatures=None
     18 )
     19 

2 frames
/usr/local/lib/python3.6/dist-packages/tensorflow_core/python/keras/saving/saving_utils.py in model_metadata(model, include_optimizer, require_config)
    180           'loss': model.loss,
    181           # pylint: disable=protected-access
--> 182           'metrics': model._compile_metrics,
    183           'weighted_metrics': model._compile_weighted_metrics,
    184           # pylint: enable=protected-access

AttributeError: 'Model' object has no attribute '_compile_metrics'

Это глубоко правильное репо: https://github.com/bedapudi6788/deepcorrect

Как сохранить модель в режиме обслуживания Tensorflow?