galmoss 2.2

A Python-based, Torch-powered tool for two-dimensional fitting of galaxy profiles. By seamlessly enabling GPU parallelization, GalMOSS meets the high computational demands of large-scale galaxy surveys.

Why use GalMOSS?

Fit 8,000 galaxies in just 10 minutes!

GalMOSS a Python-based, Torch-powered tool for two-dimensional fitting of galaxy profiles. By seamlessly enabling GPU parallelization, GalMOSS meets the high computational demands of large-scale galaxy surveys, placing galaxy profile fitting in the CSST/LSST-era. It incorporates widely used profiles such as the Sérsic, Exponential disk, Ferrer, King, Gaussian, and Moffat profiles, and allows for the easy integration of more complex models.

How to install

We provide two kinds of methods to download and install GalMOSS, pip and git.

Install via pip

pip install galmoss

Install via git

git clone https://github.com/Chenmi0619/GALMoss
cd galmoss
python setup.py install

You can access it by clicking on GitHub-GALMoss.

How to use

Fit sersic profile on single galaxy

Here, we demonstrate how to fit a single Sérsic profile to SDSS image data using the GALMoss package.

First, we need to load the necessary packages.

import Galmoss as gm

Next, we need to define the parameter objects and associate them with profile instances. The initial estimates of the galaxy parameters are provided by texttt{sextractor}. Notably, we do not include the boxiness parameter in this simple example, despite its availability within the GalMOSS framework.

# define parameter objects and profile

sersic = gm.lp.Sersic(
    cen_x=gm.p(65.43),
    cen_y=gm.p(64.95),
    pa=gm.p(-81.06, angle=True),
    axis_r=gm.p(0.64),
    eff_r=gm.p(7.58, pix_scale=0.396),
    ser_n=gm.p(1.53, log=True),
    mag=gm.p(17.68, M0=22.5)
)

The comprehensive dataset object can be formulated utilising the image sets (galaxy image, mask image, PSF image, sigma image) together with the chosen profiles.

dataset = gm.Dataset(
    galaxy_index="J162123.19+322056.4",
    image_path="./J162123.19+322056.4_image.fits",
    sigma_path="./J162123.19+322056.4_sigma.fits",
    psf_path="./J162123.19+322056.4_psf.fits",
    mask_path="./J162123.19+322056.4_mask.fits"
    mask_index=2,
    img_block_path="./test_repo",
    result_path="./test_repo"
)

dataset.define_profiles(sersic=sersic)

After initializing the hyperparameter during the fitting process, training could start. Subsequently, we run the uncertainty estimation process.

fitting = gm.Fitting(dataset=dataset,
                    batch_size=1,
                    iteration=1000)
fitting.fit()
fitting.uncertainty(method="covar_mat")

When the fitting process is completed, the fitted results and the img_blocks are saved in corresponding path.

Fit bulge+disk profile on multiple galaxies

Here, we demonstrate how to use a combination of two Sérsic profiles to make disk and bulge decomposition on SDSS image data using the GalMOSS package.

import Galmoss as gm

Upon importing the package, the subsequent step entails defining parameter objects. To ensure that the center parameter within both profiles remains the same, it suffices to specify the center parameter once and subsequently incorporate it into various profiles.

xcen = gm.p([65.97, 65.73])
ycen = gm.p([65.30, 64.81])

For a quick start, we let the disk and bulge profile share the initial value from the SExtractor, with an initial Sérsic index of 1 for the bulge component and 4 for the disk component.

bulge = gm.lp.Sersic(cen_x=xcen,
                    cen_y=ycen,
                    pa=gm.p([58.7, -8.44], angle=True),
                    axis_r=gm.p([0.75, 0.61709153]),
                    eff_r=gm.p([4.09, 18], pix_scale=0.396),
                    ser_n=gm.p([4, 4], log=True),
                    mag=gm.p([17.97, 15.6911], M0=22.5))

disk = gm.lp.Sersic(cen_x=xcen,
                    cen_y=ycen,
                    pa=gm.p([58.7, -8.44], angle=True),
                    axis_r=gm.p([0.75, 0.61709153]),
                    eff_r=gm.p([4.09, 18], pix_scale=0.396),
                    ser_n=gm.p([1, 1], log=True),
                    mag=gm.p([17.97, 15.6911], M0=22.5))

Compared to the single profile case, we only need to change the code of profile definition. We choose to use bootstrap to calculate the uncertainty here.

dataset = gm.DataSet(["J100247.00+042559.8", "J092800.99+014011.9"],
                image_path=["./J100247.00+042559.8_image.fits",
                            "./J092800.99+014011.9_image.fits"],
                sigma_path=["./J100247.00+042559.8_sigma.fits",
                            "./J092800.99+014011.9_sigma.fits"],
                psf_path=["./J100247.00+042559.8_psf.fits",
                        "./J092800.99+014011.9_psf.fits"],
                mask_path=["./J100247.00+042559.8_mask.fits",
                        "./J092800.99+014011.9_mask.fits"],
                img_block_path="./test_repo/",
                result_path="./test_repo/"
)
dataset.define_profiles(bulge=bulge, disk=disk)
fitting = gm.Fitting(dataset=dataset,
                    batch_size=1,
                    iteration=1000)
fitting.fit()
fitting.uncertainty(method="bstrap")

Requirements

numpy>=1.21.0

pandas>=1.4.4

torch>=2.0.1

astropy>=5.1

h5py>=3.7.0

torch-optimizer>=0.3.0

tqdm>=4.64.1