Embodied AI Glossary中文

Conda

Common

A widely used Python package and virtual-environment manager that isolates dependencies between projects.

Conda is an open-source package- and environment-management tool, originally released by the Anaconda company. It can create a separate environment for each project, each with its own versions of Python and libraries that don’t interfere with one another; beyond Python packages, it can also install non-Python dependencies such as the CUDA toolkit and compilers. Anaconda is the full distribution that bundles a large set of scientific-computing packages; Miniconda is a stripped-down version with just conda itself; Miniforge is the community edition, defaulting to downloads from the conda-forge channel. Dependency versions in embodied AI projects often conflict with each other (a simulator, for instance, might require a specific Python version), so most open-source install instructions start with creating a conda environment.

Exampleconda create -n lerobot python=3.10, then conda activate lerobot, then pip install -e . as the project’s instructions describe.

Also called
Anaconda, Miniconda, Miniforge, conda environment
Related
uv · Docker · PyTorch · CUDA · Package Mirror Sources in China (pip / conda / apt / Docker mirrors)
Sources
conda documentation

See it in the full glossary →