Getting started

Install

Python 3.12 or later is required. Until the first PyPI release is published, install from the canonical repository:

pip install git+https://github.com/jaxglitches/jaxglitches.git

Once a release is available:

pip install jaxglitches

For an editable development installation:

git clone https://github.com/jaxglitches/jaxglitches.git
cd jaxglitches
uv sync --locked

Build a waveform

import jax.numpy as jnp
import jaxglitches as jg

freq = jg.freq_grid(t_obs=1800.0, dt=0.5)
params = jnp.array([400.0, 1.2e-13, 0.79])
response = jg.clean_signal_f(params, freq, tdi=2, basis="XYZ")
print(response.shape)  # (1801, 3)

The parameter vector contains onset time in seconds, velocity kick in m/s, and decay time in seconds. Columns above are X, Y, Z; the default basis is AET. See conventions before comparing against data or noise models.

Run the example notebook for time-domain plots and JAX derivatives.

Optional dependencies

Extra	Purpose
notebook	Plotting, notebook execution, and Jexplore sampling
joint	Galactic-binary responses through jaxgb
wdm	Wavelet transforms through wdm-transform
simulation	lisainstrument and pytdi validation tools
gpu	JAX CUDA 12 support on Linux x86_64
gpu-cuda13	JAX CUDA 13 support on Linux x86_64

For the joint glitch/binary notebooks:

uv sync --locked --extra notebook --extra joint --extra wdm

GPU extras require a compatible NVIDIA driver. Verify the available devices with python -c "import jax; print(jax.devices())" after installation.