NVIDIA Inference Xfer Library (NIXL) is targeted for accelerating point to point communications in AI inference frameworks such as NVIDIA Dynamo, while providing an abstraction over various types of memory (e.g., CPU and GPU) and storage (e.g., file, block and object store) through a modular plug-in architecture.
$ sudo apt install build-essential cmake pkg-config
$ sudo dnf install gcc-c++ cmake pkg-config
$ pip3 install meson ninja pybind11
NIXL was tested with UCX version 1.18.0.
$ wget https://git
7812
hub.com/openucx/ucx/releases/download/v1.18.0/ucx-1.18.0.tar.gz
$ tar xzf ucx-1.18.0.tar.gz
$ cd ucx-1.18.0
$ ./contrib/configure-release --prefix=<PATH_TO_INSTALL>/install
$ make -j8
$ make install
$ meson setup <name_of_build_dir>
$ cd <name_of_build_dir>
$ ninja
$ ninja-install
The pybind11 bindings for the public facing NIXL API are available in src/bindings/python. These bindings implement the headers in the src/api/cpp directory.
The preferred way is to build it through meson-python, which will just let it be installed with pip. This can be done from the root nixl directory:
$pip install .
To build the docker container, first clone the current repository. Also make sure you are able to pull docker images to your machine before attempting to build the container.
Run the following from the root folder of the cloned NIXL repository:
$ docker build -t nixl-container -f contrib/Dockerfile .