deepof.model_utils.find_learning_rate

deepof.model_utils.find_learning_rate(model, data, epochs=1, batch_size=32, min_rate=1e-08, max_rate=0.1)

Train the provided model for an epoch with an exponentially increasing learning rate.

Parameters:

  • model (tf.keras.Model) – model to train

  • data (tuple) – training data

  • epochs (int) – number of epochs to train the model for

  • batch_size (int) – batch size to use for training

  • min_rate (float) – minimum learning rate to consider

  • max_rate (float) – maximum learning rate to consider

Returns:

learning rate that resulted in the lowest loss

Return type:

float