Install
Prerequisites
Before installing NVIDIA AITune, make sure your system meets these requirements:
- Operating System: Linux (Ubuntu 22.04+ recommended)
- Python: Version
3.11or newer - PyTorch: Version
2.8or newer - TensorRT: Version
10.3or higher (for TensorRT backend) - NVIDIA GPU: Required for GPU-accelerated tuning
Installing from PyPI (Recommended)
You can install NVIDIA AITune from PyPI.
Pinning to a specific PyTorch version
If you need to pin AITune’s dependencies (torch, torch-tensorrt, torchao) to a specific PyTorch minor version, use one of the version extras:
Selecting a CUDA version
To target a specific CUDA version, add the matching PyTorch index:
ONNX Runtime CUDA version
Select the CUDA 13 extra when installing from a source checkout with UV:
Use the onnxruntime-gpu-cuda12 extra for the CUDA 12 PyPI build:
pip uses PyPI and therefore installs CUDA 12. Use the same extra explicitly:
CUDA 13 with pip needs an explicit ORT reinstall from the CUDA 13 feed:
If ORT is already installed, use UV’s targeted reinstall option with the CUDA 13 extra:
NGC container install
You can use NGC PyTorch containers, which include the required NVIDIA dependencies:
Inside an NGC container, use pip and the package name:
Installing from Source
or use editable mode for development: