You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Our current loss functions are implemented using tf.keras.backend. This throws an error when we use more memory-efficient data types like bfloat16 or other mixed-precision strategies. We could potentially reimplement the loss functions using low-level Tensorflow function (without relying on Keras backend functions)
The text was updated successfully, but these errors were encountered:
Our current loss functions are implemented using tf.keras.backend. This throws an error when we use more memory-efficient data types like bfloat16 or other mixed-precision strategies. We could potentially reimplement the loss functions using low-level Tensorflow function (without relying on Keras backend functions)
The text was updated successfully, but these errors were encountered: