Installation

DataEval is a library that offers powerful metric classes and dataset analysis functions using NumPy and PyTorch as the primary backends.

Supported Python Versions

We currently support Python versions >= 3.10

PyTorch Dependency

DataEval requires PyTorch to be installed. By default, pip install dataeval pulls PyTorch from PyPI, which includes CUDA support on Linux.

To install a specific PyTorch variant, use --extra-index-url to point pip at the appropriate PyTorch wheel index:

# CPU only
pip install dataeval --extra-index-url https://download.pytorch.org/whl/cpu

# CUDA 11.8
pip install dataeval --extra-index-url https://download.pytorch.org/whl/cu118

# CUDA 12.8
pip install dataeval --extra-index-url https://download.pytorch.org/whl/cu128

See the PyTorch installation guide for all available PyTorch installation options.

Note: When installing from source using uv, you can use extras to specify PyTorch versions (e.g., --extra cpu, --extra cu118, --extra cu128). See the source installation instructions below for details.

Installing DataEval

Now that you have chosen which DataEval to install, the following methods will show you how to install using your preferred method.

Installing from pip

pip install dataeval

Installing from conda

conda install -c conda-forge dataeval

To install DataEval from source locally on Ubuntu using poetry, begin by ensuring poetry is installed in your Python environment.

    pip install poetry

Pull the source down and change to the DataEval project directory.

    git clone https://github.com/aria-ml/dataeval.git
    cd dataeval

Install DataEval

    poetry install

Now that DataEval is installed, you can run commands in the Poetry virtual environment by prefixing shell commands with poetry run, or activate the virtual environment directly in the shell.

    poetry env activate

To install DataEval from source locally on Ubuntu, you will need uv for Python environment management.

Pull the source down and change to the DataEval project directory.

    git clone https://github.com/aria-ml/dataeval.git
    cd dataeval

Install DataEval with development dependencies.

    uv sync

Optionally, you can specify the version of Python and PyTorch CPU/CUDA support (cpu, cu118, cu128) using -p and –extra respectively.

For example, the following command installs DataEval in a Python 3.11 environment using only PyTorch with CPU support, and no development dependencies:

    uv sync -p 3.11 --extra cpu --no-default-groups

Now that DataEval is installed, you can run commands in the uv virtual environment by prefixing shell commands with uv run, or activate the virtual environment directly in the shell.

    source .venv/bin/activate