NVIDIA NVJITLink is a Python package that distributes the native NVIDIA NVJITLink shared library (libnvjitlink). This library provides JIT (Just-In-Time) compilation services for NVIDIA GPU code, acting as a low-level component within the NVIDIA ecosystem. It's primarily used as a backend dependency by higher-level Python libraries and frameworks (e.g., Numba, CuPy, PyTorch with NVFuser) that need to compile GPU kernels at runtime. Its version is often tightly coupled with the NVIDIA CUDA Toolkit version. The current version is 13.2.51, and it's updated in conjunction with CUDA Toolkit releases.
pip install nvidia-nvjitlinkVerified import paths — ran on the pinned version, not inferred.
This quickstart demonstrates how to import the `nvidia_nvjitlink` package and access its `__version__` attribute, which is the most common direct interaction for end-users. It verifies the package's presence and version, confirming it's available for other GPU-accelerated libraries that depend on it.
Do not attempt to write GPU code directly using `nvidia-nvjitlink`. Instead, use libraries built on top of it, such as Numba for JIT compilation, or PyTorch/CuPy for array-based GPU computation.
Ensure that the `nvidia-nvjitlink` version is compatible with your CUDA driver and toolkit. When using dependent libraries (e.g., Numba, CuPy), always check their recommended installation procedures and dependencies, as they often handle `nvidia-nvjitlink` compatibility implicitly.
Prioritize installing the higher-level GPU library you intend to use. Allow it to manage `nvidia-nvjitlink` as its dependency. Only install `nvidia-nvjitlink` directly if explicitly instructed or if you are developing a low-level library that specifically requires direct access to `libnvjitlink`.
No dependency data recorded yet.