Installing from PyPI

Nsight Python can be installed directly from the Python Package Index (PyPI).

Basic Installation

For most users, a basic installation is sufficient:

pip install nsight-python

This installation works with both CUDA Toolkit 12 and CUDA Toolkit 13, and provides all core Nsight Python features.

Optional: Installing with cuda-core Support

If you want to use the ignore_failures feature in nsight.annotate, you need to install the cuda-core package. This is an optional dependency that enables enhanced error handling within annotated regions.

For CUDA Toolkit 12:

pip install nsight-python[cu12]

For CUDA Toolkit 13:

pip install nsight-python[cu13]

Optional: Installing with CUPTI Support

If you want to use the experimental CUPTI backend with nsight.analyze.kernel, you need to install the cupti-python package. This is an optional dependency that enables low-overhead collection of kernel time durations.

pip install nsight-python[cupti]

Note

The CUPTI backend is not supported on Windows.