Installation
Overview
Arctic Platform can be installed from PyPI for the latest stable release or from source for the latest development version. Installation in editable mode is recommended for active development.
Installation
Arctic Platform supports two installation methods: from PyPI for stable releases, or from source for the latest development version.
From PyPI
Install the latest released version and its dependencies from PyPI:
pip install arctic-platform
Note: This project is under active development. For the latest features and improvements, consider installing from source instead.
From Source (Git)
To install the latest development version or to contribute to the project, clone the repository and install it in editable mode:
git clone https://github.com/Snowflake-AI-Research/Arctic-Platform.git
cd Arctic-Platform
pip install -e .[rl]
The [rl] extra installs dependencies required for Reinforcement Learning components, including:
liger-kernel>=0.8.0raytensordictuvicornvllm==0.18.0
Installation Extras
The package supports several optional dependency groups:
| Extra | Purpose |
|---|---|
[rl] | Reinforcement Learning components |
[dev] | Development tools (formatting and testing) |
[testing] | Testing frameworks and utilities |
[formatting] | Code formatting tools (pre-commit, autoflake) |
[all] | All optional dependencies |
Install a specific extra with:
pip install -e .[dev]
pip install -e .[testing]
pip install -e .[all]
Requirements
- Python: 3.10 or later (tested on 3.10, 3.11, and 3.12)
- Core dependencies: torch, transformers (>=4.57.0,<5), deepspeed (>=0.19.2), peft, and others
- GPU support: NVIDIA GPUs recommended for RL training and inference
For a complete list of core dependencies, see pyproject.toml.
Next Steps
After installation, follow the recipes to get started with training models using Arctic Platform.