Skip to content

rickardp/bitsandbytes

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

bitsandbytes

The bitsandbytes library is a lightweight Python wrapper around CUDA custom functions, in particular 8-bit optimizers, matrix multiplication (LLM.int8()), and 8 & 4-bit quantization functions.

The library includes quantization primitives for 8-bit & 4-bit operations, through bitsandbytes.nn.Linear8bitLt and bitsandbytes.nn.Linear4bit and 8-bit optimizers through bitsandbytes.optim module.

There are ongoing efforts to support further hardware backends, i.e. Intel CPU + GPU, AMD GPU, Apple Silicon. Windows support is quite far along and is on its way as well.

Please head to the official documentation page:

https://huggingface.co/docs/bitsandbytes/main

License

The majority of bitsandbytes is licensed under MIT, however small portions of the project are available under separate license terms, as the parts adapted from Pytorch are licensed under the BSD license.

We thank Fabio Cannizzo for his work on FastBinarySearch which we use for CPU quantization.

About

8-bit CUDA functions for PyTorch

Resources

License

Code of conduct

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Python 58.5%
  • Cuda 27.0%
  • C++ 11.7%
  • Shell 1.3%
  • CMake 0.9%
  • Metal 0.3%
  • Other 0.3%