UnknownError: Graph execution error: JIT compilation failed.

[latheeshvm]

TensorFlow certification requires : tensorflow==2.9.0

sentence_encoder_layer = hub.KerasLayer( "https://tfhub.dev/google/universal-sentence-encoder/4", input_shape=[], dtype=tf.string, trainable=False, name="USE")

this is throwing errors like this

JIT compilation failed.
	 [[{{node EncoderDNN/EmbeddingLookup/EmbeddingLookupUnique/embedding_lookup/mod}}]] [Op:__inference_train_function_35744]

Complete model

import tensorflow_hub as hub

sentence_encoder_layer = hub.KerasLayer(
    "https://tfhub.dev/google/universal-sentence-encoder/4", input_shape=[], dtype=tf.string, trainable=False, name="USE")

model_6 = tf.keras.Sequential([
    sentence_encoder_layer,
    # tf.keras.layers.Dense(64, activation="relu"),
    tf.keras.layers.Dense(1, activation="sigmoid")
])
model_6.compile(loss="binary_crossentropy",
                optimizer=tf.keras.optimizers.Adam(), metrics=["accuracy"])

model_6_history = model_6.fit(train_sentences, train_labels, epochs=5, validation_data=(
    val_sentences, val_labels), callbacks=[create_tensorboard_callback("model_logs", "model_6_transferlearning")])

process_the_model_output(tf, model_5, val_sentences,
                         val_labels, calculate_results)


Please suggest a fix

[latheeshvm]

Solved it. for TensorFlow 2.9.0 -> it needs a cuda 11.2