GPyTorch-based models¶

A number of models implemented in heron make use of pytorch and the GPR library built atop it, gpytorch. These models can be used on both CPU and GPU hardware.

All of these models are contained within the heron.models.torchbased module.

HeronCUDA : A spinning, NR-trained, GPU-capable surrogate model¶

Training data	GPR Technique	Model type	Spinning	Higher modes
NR: Georgia Tech	Exact, LOVE, CUDA	BBH	Fully	No

The model is trained on numerical relativity waveforms produced by the Centre for Relativistic Astrophysics at Georgia Tech, and uses exact scalable GPR techniques implemented by GPyTorch.

class heron.models.torchbased.HeronCUDA(device=device(type='cuda'))[source]¶

A GPR BBH waveform model which is capable of using CUDA resources.

Methods

`bilby`(self, time, mass_1, mass_2, …)	Return a waveform from the GPR in a format expected by the Bilby ecosystem
`build`(self)	Right now this isn’t need by this method
`distribution`(self, times, p[, samples])	Return a number of sample waveforms from the GPR distribution.
`eval`(self)	Prepare the model to be evaluated.
`frequency_domain_waveform`())	Return the frequency domain waveform.
`mean`(self, times, p[, covariance])	Provide the mean waveform and its variance.
`plot_td`([f])	Plot the timedomain waveform.
`time_domain_waveform`())	Return the timedomain waveform.