CODEWITHSUNDEEP
tensorflowtensorboardtensorflow-estimator
bluesummers
bluesummers

Reputation: 12657

Disable TensorBoard logging on tf.Estimator methods

Is there a way disable the automatic TensorBoard logging when using the tf.estimator.Estimator class?

# Classifier
classifier = tf.estimator.Estimator(
    model_fn=lambda features, labels, mode, params: model(features, labels, mode, params, word_embeddings),
    params=params,
    model_dir=str(MODEL_DIR),
    tensorboard=Fasle)  # <---- Something like that?

for _ in range(params['epochs']):
    classifier.train(input_fn=lambda: input_fn(generator, example_generator._data['train'] ,batch_size=params['batch_size']))
    classifier.evaluate(input_fn=lambda: input_fn(generator, example_generator._data['validation'], batch_size=params['batch_size']))

I read through the tf.estimator.RunConfig and couldn't find a solution.

Upvotes: 0

Views: 413

Answers (1)

Arsey
Arsey

Reputation: 281

Instantiating RunConfig for Estimator with save_summary_steps=None should fix this

Upvotes: 2

Related Questions