W&B Python Library wandb Version 0.12.16 Released
Version 0.12.16 of the wandb Python library has released with support for Keras SavedModel format and more.
Created on May 5|Last edited on May 6
Comment
Version 0.12.16 of W&B's very own Python library wandb has released with a nice collection of improvements and bug fixes.
The highlight of this update is the support for the TensorFlow/Keras SavedModel format. Before this update, using WandbCallback would save the model in the HDF5 format which is not the recommended format when using Keras since it cannot save subclasses models or those with custom layers, however now it will additionally save the model in the SavedModel format as a W&B Artifact.
A few additional changes of note are that logging support for Jax's bfloat16 datatype is now added, excessive table size limits are now handled in a better way, a timeout setting has been added to wandb.init(), and much more.
Find out more
Add a comment
Iterate on AI agents and models faster. Try Weights & Biases today.