Skip to content

Instantly share code, notes, and snippets.

@eerkaijun

eerkaijun/fine_tune.py

Created August 14, 2020 14:00
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 eerkaijun/ec135a44951d3c144468d0deb6cd1d7b to your computer and use it in GitHub Desktop.
Save eerkaijun/ec135a44951d3c144468d0deb6cd1d7b to your computer and use it in GitHub Desktop.
Download ZIP
Fine tune BERT model
Raw
fine_tune.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')
model.compile(optimizer=optimizer, loss=loss, metrics=[metric])
model.fit(train_data, epochs=7, validation_data=validation_data)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment