tricks.md

Tricks for TensorFlow

• tf.keras.utils.disable_interactive_logging
• tf.keras.callbacks.ModelCheckpoint
• map interleave and flat_map

num_of_cores = multiprocessing.cpu_count() # num of available cpu cores
mapped_data = data.map(function, num_parallel_calls = num_of_cores)

Documentation

matteo_without_skip_connection

best found - loss: 0.0108 - accuracy: 0.9962 final - loss: 0.0306 - accuracy: 0.9879

matteo_small_latent_space_Adam

best yet (2nd epoch) - loss: 0.0367 - accuracy: 0.9832 final -

matteo_small_latent_space_SGD (learning rate 0.00001)

• will not converge better than Adam best yet final

matteo_small_latent_spaceSGDs (learning rate 0.000001)

• will not converge better than Adam