Subclassing tf.keras objects (Layer, Model, Metric...), together with eager execution & the GradientTape, bridges the gap from the lowest-level APIs to the highest-level features like Sequential and fit(). Full flexibility + productivity. Check it out: https://www.tensorflow.org/alpha/guide/keras/custom_layers_and_models …https://twitter.com/random_forests/status/1103428693660037121 …
-
Show this thread
-
This Tweet is unavailable.
With subclassing you have to separate state-building (in __init__ or build) and execution (in call). But if you use the functional API (where you can use custom layers too) you can do both in a single step.
3:24 PM - 6 Mar 2019
0 replies
0 retweets
3 likes
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.