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
I've been trying to migrate my TF=2.15 and TFP=0.23 code to Keras 3 with JAX (+TF=2.18 tf.data) and stumbled upon Distrax which seems like an elegant drop-in solution :)
However, I've noticed that importing Distrax imports TFP via from tensorflow_probability.substrates import jax as tfp which then asks for tf-keras, i.e. Keras 2.
Failed to import TF-Keras.
Please note that TF-Keras is not installed by default when you install TensorFlow Probability.
This is so that JAX-only users do not have to install TensorFlow or TF-Keras.
To use TensorFlow Probability with TensorFlow, please install the tf-keras or tf-keras-nightly package.
This can be be done through installing the tensorflow-probability[tf] extra.
Does that imply that Distrax is therefore incompatible with Keras 3?
The text was updated successfully, but these errors were encountered:
I've been trying to migrate my TF=2.15 and TFP=0.23 code to Keras 3 with JAX (+TF=2.18 tf.data) and stumbled upon Distrax which seems like an elegant drop-in solution :)
However, I've noticed that importing Distrax imports TFP via
from tensorflow_probability.substrates import jax as tfp
which then asks for tf-keras, i.e. Keras 2.Does that imply that Distrax is therefore incompatible with Keras 3?
The text was updated successfully, but these errors were encountered: