You can add location information to your Tweets, such as your city or precise location, from the web and via third-party applications. You always have the option to delete your Tweet location history. Learn more
This how to train a Keras model in mixed precision (float32 weights & float16 computation). After this line, you can proceed to build a model and call `fit()` -- no other requirements. It provides a nice speed boost, esp. for image models Detailed guide: https://www.tensorflow.org/guide/mixed_precision …pic.twitter.com/rZyg13BSOa
The API also provides a simple way to configure the dtype policy of each individual layer in a model (just pass a `dtype` argument to your layer). This can be handy in cases where you have a layer that has a tendency to produce NaNs in mixed precision.
Simplicity 
It gets super slow when the model contains tf.image.resize op with bilinear interpolation. ~6x-10x slower
I prefer being able to use the sweet, sweet SWA optimizer in TF Addons over mixed precision. And coworker mentioned a performance hit with mixed. Dunno if that's because we use mish instead of relu.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
