tensorflow: Cannot convert model containing categorical_column_with_vocabulary_list op

System information

• OS Platform and Distribution (e.g., Linux Ubuntu 16.04):CentOS
• TensorFlow installed from (source or binary): binary
• TensorFlow version (or github SHA if from source): 2.2.0rc0

Command used to run the converter or code if you’re using the Python API If possible, please share a link to Colab/Jupyter/any notebook.

import tensorflow as tf
import os

model_dir = "models/feature_column_example"
category = tf.constant(["A", "B", "A", "C", "C", "A"])
label = tf.constant([1, 0, 1, 0, 0, 0])

ds = tf.data.Dataset.from_tensor_slices(({"category": category}, label))
ds = ds.batch(2)

fc_category = tf.feature_column.indicator_column(
    tf.feature_column.categorical_column_with_vocabulary_list(
        "category", vocabulary_list=["A", "B", "C"]
    )
)
feature_layer = tf.keras.layers.DenseFeatures([fc_category])

model = tf.keras.Sequential(
    [
        feature_layer,
        tf.keras.layers.Dense(10, activation="relu"),
        tf.keras.layers.Dense(1, activation="sigmoid"),
    ]
)
model.compile(
    optimizer="adam", loss="binary_crossentropy", metrics=["accuracy"]
)

model.fit(ds, epochs=2)
converter = tf.lite.TFLiteConverter.from_keras_model(model)
converter.allow_custom_ops = True
# converter.experimental_new_converter = True
# converter.experimental_new_quantizer = True

# Convert the model.
tflite_model = converter.convert()
open(os.path.join(model_dir, "output.tflite"), "wb").write(tflite_model)

The output from the converter invocation

Cannot convert a Tensor of dtype resource to a NumPy array.

Also, please include a link to the saved model or GraphDef

saved_model_cli show --dir models/feature_column_example/ --tag_set serve --signature_def serving_default

The given SavedModel SignatureDef contains the following input(s):
  inputs['category'] tensor_info:
      dtype: DT_STRING
      shape: (-1, 1)
      name: serving_default_category:0
The given SavedModel SignatureDef contains the following output(s):
  outputs['output_1'] tensor_info:
      dtype: DT_FLOAT
      shape: (-1, 1)
      name: StatefulPartitionedCall_7:0
Method name is: tensorflow/serving/predict

Failure details Cannot convert a Tensor of dtype resource to a NumPy array.

According to my analysis, this might be caused by some HashTable Ops, which create table handles. And my additional question is: whether tfliteconverter could convert model contains ops of initialize hashtableV2 and LookupTableImportV2? Thank you.

Any other info / logs: Full logs

Traceback (most recent call last):
  File "feature_column_example.py", line 62, in <module>
    export_keras_hashtable_model()
  File "feature_column_example.py", line 58, in export_keras_hashtable_model
    tflite_model = converter.convert()
  File "/root/tf2.2/lib/python3.6/site-packages/tensorflow/lite/python/lite.py", line 464, in convert
    self._funcs[0], lower_control_flow=False))
  File "/root/tf2.2/lib/python3.6/site-packages/tensorflow/python/framework/convert_to_constants.py", line 706, in convert_variables_to_constants_v2_as_graph
    func, lower_control_flow, aggressive_inlining)
  File "/root/tf2.2/lib/python3.6/site-packages/tensorflow/python/framework/convert_to_constants.py", line 457, in _convert_variables_to_constants_v2_impl
    tensor_data = _get_tensor_data(func)
  File "/root/tf2.2/lib/python3.6/site-packages/tensorflow/python/framework/convert_to_constants.py", line 217, in _get_tensor_data
    data = val_tensor.numpy()
  File "/root/tf2.2/lib/python3.6/site-packages/tensorflow/python/framework/ops.py", line 961, in numpy
    maybe_arr = self._numpy()  # pylint: disable=protected-access
  File "/root/tf2.2/lib/python3.6/site-packages/tensorflow/python/framework/ops.py", line 929, in _numpy
    six.raise_from(core._status_to_exception(e.code, e.message), None)
  File "<string>", line 3, in raise_from
tensorflow.python.framework.errors_impl.InvalidArgumentError: Cannot convert a Tensor of dtype resource to a NumPy array.