request help with error: AttributeError: 'SymbolicTensor' object has no attribute 'numpy'

[Tex Cowboy]

Hi team,

I trying to run the following sample code  from https://keras.io/keras_hub/api/models/deberta_v3/deberta_v3_masked_lm/

import keras_hub

# KERAS_BACKEND was set to tensorflow or jax

def test_running_deberta_lm():

    features = ["The quick brown fox jumped.", "I forgot my homework."]

    masked_lm = keras_hub.models.DebertaV3MaskedLM.from_preset('deberta_v3_base_en')

    masked_lm.fit(x=features, batch_size=2)

    masked_lm.summary()

if __name__ == '__main__':

    test_running_deberta_lm()

I am using Python 3.11 venv and have installed the latest requirements-jax-cuda.txt from keras-hub.

However, I am getting the following error when running masked_lm.fit(...)

sentence_piece_tokenizer.py", line 172, in vocabulary_size
        return int(self._sentence_piece.vocab_size().numpy())

    AttributeError: 'SymbolicTensor' object has no attribute 'numpy'

I appreciate any tips!

Not sure if this had to do with running eager execution, but this version is set up to run jax...

Thanks in advance