Skip to content

Instantly share code, notes, and snippets.

@LysandreJik

LysandreJik/keras_fit.py Secret

Created October 8, 2019 02:50
Show Gist options
  • Star 0 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save LysandreJik/b295e94bbb569c7630cf466f998c9b29 to your computer and use it in GitHub Desktop.
Save LysandreJik/b295e94bbb569c7630cf466f998c9b29 to your computer and use it in GitHub Desktop.
Download ZIP
Training with Keras' fit
Raw
keras_fit.py
optimizer = tf.keras.optimizers.Adam(learning_rate=3e-5, epsilon=1e-08, clipnorm=1.0)
loss = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True)
metric = tf.keras.metrics.SparseCategoricalAccuracy('accuracy')
bert_model.compile(optimizer=optimizer, loss=loss, metrics=[metric])
bert_history = bert_model.fit(
bert_train_dataset,
epochs=2,
steps_per_epoch=115,
validation_data=bert_validation_dataset,
validation_steps=7
)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment