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.
This fork is developed for ROCm and updates are pushed into upstream. Users can use either of these to run bitsandbytes on AMD GPUs.
**Note: The default branch of this fork is switched from this branch to the new rocm_enabled branch. **
Installation for ROCm:
For latest develop version:
git clone --recurse https://github.com/ROCm/bitsandbytes
cd bitsandbytes
git checkout rocm_enabled
cmake -DCOMPUTE_BACKEND=hip -S . #Use -DBNB_ROCM_ARCH="gfx90a;gfx942" to target specific gpu arch
make
pip install .For more details, please head to the official documentation page:
https://huggingface.co/docs/bitsandbytes/main
bitsandbytes is MIT licensed.
We thank Fabio Cannizzo for his work on FastBinarySearch which we use for CPU quantization.