A fast elliptical accretion disk modeling engine built on JAX and NumPyro.

feadme implements the elliptical accretion disk model described in Eracleous et al. (1995) and fits it to double-peaked emission line profiles via Bayesian inference.

Features

• Fast: JAX-accelerated disk integration with JIT compilation and GPU support.
• Bayesian: NUTS-based posterior sampling via NumPyro with full uncertainty quantification.
• Flexible: JSON-driven model templates supporting multiple disk and line components with shared parameters.

Installation

pip install feadme         # CPU
pip install feadme[gpu]    # GPU (CUDA 12)

From source with uv:

git clone https://github.com/nmearl/feadme.git
cd feadme
uv sync --group dev

From source without uv:

git clone https://github.com/nmearl/feadme.git
cd feadme
pip install -e ".[dev]"

Quickstart

feadme run --template-path my_template.json --data-path my_data.csv

See the documentation for the full CLI reference, data format, and template specification.

Contributing

Bug reports, feature requests, and pull requests are welcome on the GitHub repository.