This document provides step-by-step instructions to set up and install the nvdlfw_inspect tool. The recommended setup is to use the Nvidia PyTorch Docker container as the base environment. Additional instructions include installing directly via GitHub, setting up a development environment, running tests, and building the wheel.
You can install nvdlfw_inspect directly from its GitHub repository using pip, assuming you already have the pytorch docker container running
pip install git+https://github.com/NVIDIA/nvidia-dlfw-inspect
You can also clone the repostory, run pip install . / make install or run in the dev mode pip install .[dev]. You can add the repository path to $PYTHONPATH env variable to access the tool if you dont want to install it.
The installation can be verified by running python3 -c "import nvdlfw_inspect".
- Pull the NVIDIA PyTorch Docker Image:
docker pull nvcr.io/nvidia/pytorch:24.12-py3
- Build the Docker Image:
Run the following command from the root of the project directory (where the
Makefileis located):
make docker-build
- Run the Docker Container: Launch an interactive container with GPU support:
make docker-run
For local development (when not using a pytorch container), you can set up the local development environment by running: make dev-local and then source source dev_local_venv/bin/activate
To contribute or modify the tool, you can also set up a development environment: make dev-env and activate the virtual environment source dev_venv/bin/activate. This can only be used for running the pre-commit checks: make check
To ensure everything is working as expected, you can run tests using pytest: make test
To create a distributable .whl file for nvdlfw_inspect:
- Run the build script:
./scripts/build_wheels.sh
The resulting wheel file will be located in the dist/ directory.