jax 0.4.26

Differentiate, compile, and transform Numpy code.

Stars: 27720, Watchers: 27720, Forks: 2527, Open Issues: 1665

The google/jax repo was created 5 years ago and the last code push was 16 minutes ago.
The project is extremely popular with a mindblowing 27720 github stars!

How to Install jax

You can install jax using pip

pip install jax

or add it to a project with poetry

poetry add jax

Package Details

Author

JAX team

License

Apache-2.0

Homepage

https://github.com/google/jax

PyPi:

https://pypi.org/project/jax/

GitHub Repo:

https://github.com/google/jax

Classifiers

Related Packages

Errors

A list of common jax errors.

Code Examples

Here are some jax code examples and snippets.

GitHub Issues

The jax package has 1665 open issues on GitHub

• Incorrect gradient of function with segment_prod
• Allow comparing NamedShape to None
• Implement jax2tf scatter_* ops with no XLA
• unjitted_loop_body is not constantly re-compiled in the JIT tutorial
• Add GDA to the API pages
• [sparse] accept nse argument to sparse.empty()
• Efficient argmin/argmax for bool type arrays
• Fix auto-generated docstrings for JIT-compiled functions
• jax.numpy.nanpercentile with axis as tuple
• introduce custom_batching.sequential_vmap
• Jitted function sometimes doesn't distinguish static arguments with different type
• Jax profiler won't work with Cuda 11.5
• jnp.[nan]argmin/max: implement keepdims
• jax.numpy: add where and initial arguments to nan reductions
• jax2tf no_xla implementation for scatter_*, advice requested

See more issues on GitHub