DeepSpeed-Kernels is a backend library that is used to power DeepSpeed-FastGen to achieve accelerated text-generation inference through DeepSpeed-MII. This library is not intended to be an independent user package, but is open-source to benefit the community and show how DeepSpeed is accelerating text-generation.
The resulting binaries that are compiled from this repo and included in the PyPI release are torch and python agnostic, this allows the core backend to be as portable as possible and leaves the task of compiling torch and python bindings to DeepSpeed itself using it's JIT op builder.
If your environment supports it you can quickly install DeepSpeed-Kernels from PyPI (see below). We've tested the portability of the PyPI release on A100, A6000, and H100.
The release on PyPI should work with the following assumptions about your environment:
- NVIDIA GPU(s) with compute capability of: 8.0, 8.6, 8.9, 9.0
- CUDA 11.6+
- Ubuntu 20+
pip install deepspeed-kernels
If the PyPI release does not work for you we recommend installing from source which can take several minutes:
pip install -v .You can create a pre-compiled portable wheel that supports different CUDA architectures via the CUDA_ARCH_LIST environment variable. By default the kernels will be compiled using the native compute capability. If you want to compile for more than one you can set the CUDA_ARCH_LIST environment variable. We currently only support Ampere and above architectures (i.e., 8.0+). See example below to build for GPUs like A100 and A6000:
CUDA_ARCH_LIST="80;86;89;90" python -m build --wheel