flax 0.12.4

Flax: A neural network library for JAX designed for flexibility

Stars: 7078, Watchers: 7078, Forks: 788, Open Issues: 464

The google/flax repo was created 6 years ago and the last code push was 13 hours ago.
The project is extremely popular with a mindblowing 7078 github stars!

How to Install flax

You can install flax using pip

pip install flax

or add it to a project with poetry

poetry add flax

Package Details

Author

None

License

None

Homepage

None

PyPi:

https://pypi.org/project/flax/

GitHub Repo:

https://github.com/google/flax

Classifiers

• Scientific/Engineering/Artificial Intelligence

Related Packages

Errors

A list of common flax errors.

Code Examples

Here are some flax code examples and snippets.

GitHub Issues

The flax package has 464 open issues on GitHub

• feat(nnx): Expose GQA support (num_key_value_heads) in MultiHeadAttention
• optax docs test
• allow nnx standalone import
• Improvement on the Training API
• This CL relaxes the input type hints to resolve an invariance issue with Python type checkers (Pyright/MyPy).
• nnx.Optimizer.update should return the updates
• feat(nnx): add Grouped Query Attention (GQA) support
• feat(nnx): add preferred_element_type to attention.py and recurrent.py, out_sharding to recurrent.py
• set_mode guide
• Support QGA in nnx dot_product_attention
• Adding transformer encoder and decoder layers to flax source as in pytorch
• docs(nnx): standardize 'rngs' docstring description
• [graphcast] Prepare for jax_pmap_shmap_merge=True.
• [docs]: Improve docs for NNX API Reference
• Add out_sharding docstrings to linear layer call methods

See more issues on GitHub