Metadata-Version: 2.1
Name: FJFormer
Version: 0.0.25
Home-page: https://github.com/erfanzar/FJFormer
Author: Erfan Zare Chavoshi
Author-email: erfanzare82@yahoo.com
License: Apache License 2.0
Classifier: Programming Language :: Python :: 3
Classifier: License :: OSI Approved :: MIT License
Classifier: Operating System :: OS Independent
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.8
Classifier: Programming Language :: Python :: 3.9
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Requires-Python: >=3.8
Description-Content-Type: text/markdown
License-File: LICENSE

# FJFormer

Embark on a journey of paralleled/unparalleled computational prowess with FJFormer - an arsenal of custom Jax Flax
Functions and Utils that elevate your AI endeavors to new heights!

## Overview

FJFormer is a collection of functions and utilities that can help with various tasks when using Flax and JAX. It
includes
checkpoint savers, partitioning tools, and other helpful functions.
The goal of FJFormer is to make your life easier when working with Flax and JAX. Whether you are training a new model,
fine-tuning an existing one, or just exploring the capabilities of these powerful frameworks, FJFormer offers

- FlashAttention on `TPU/GPU` 🧬
- BITComputations for 8,6,4 BIT Flax Models
- Smart Dataset Loading
- Built-in functions and Loss functions
- GPU-Pallas triton like implementation of `Softmax`, `FlashAttention`, `RMSNorm`, `LayerNorm`
- Distributed and sharding Model Loaders and Checkpoint Savers
- Monitoring Utils for *TPU/GPU/CPU* memory `foot-print`
- Special Optimizers with schedulers and Easy to Use
- Partitioning Utils

and A lot of these features are fully documented so i gusse FJFormer has something
to offer, and it's not just a Computation BackEnd for [EasyDel](https://github.com/erfanzar/EasyDel).

checkout for documentations [here](https://erfanzar.github.io/FJFormer/).

## Contributing

FJFormer is an open-source project, and contributions are always welcome! If you have a feature request, bug report, or
just want to help out with development, please check out our GitHub repository and feel free to submit a pull request or
open an issue.

Thank you for using FJFormer, and happy training!

