Catalogue Fitting¶

Synference has a method specifically for fitting catalogues of sources - fit_catalogue.

This is designed to be a flexible and fast way to fit your full catalogue.

It can handle:

Transforming observations to match the expected model features.
Prior predictive checks/outlier detection to remove sources that are unlikely to be well modelled.
Optional imputation of missing data.
Rapid posterior inference using trained model.
Optional SED recovery (see this notebook for more details).
Returning structured results.

[1]:

import os

import numpy as np
from IPython.display import display
from synthesizer import get_grids_dir
from unyt import Jy

from synference import SBI_Fitter, load_unc_model_from_hdf5, test_data_dir

print(get_grids_dir())

available_grids = os.listdir(get_grids_dir())
if "test_grid.hdf5" not in available_grids:
    cmd = f"synthesizer-download --test-grids --destination {get_grids_dir()}"
    os.system(cmd)

library_path = os.path.join(get_grids_dir(), "test_grid.hdf5")

/opt/hostedtoolcache/Python/3.10.20/x64/lib/python3.10/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html
  from .autonotebook import tqdm as notebook_tqdm

/home/runner/.local/share/Synthesizer/grids

We’ll use a subset of the JADES spectroscopic catalogue used in the synference paper for this example.

[2]:

from astropy.table import Table

cat = Table.read(f"{test_data_dir}/jades_spec_catalogue_subset.fits")

The first step is the same as our SED recovery example - we will load our trained model and the noise model used to train the model. We need the noise model to transform our observations to match the model features.

[3]:

library_path = (
    f"{test_data_dir}/grid_BPASS_Chab_DenseBasis_SFH_0.01_z_14_logN_2.7_Calzetti_v3_multinode.hdf5"  # noqa: E501
)

fitter = SBI_Fitter.load_saved_model(
    model_file=f"{test_data_dir}", library_path=library_path, device="cpu"
)

nm_path = f"{test_data_dir}/BPASS_DenseBasis_v4_final_nsf_0_params_empirical_noise_models.h5"
noise_models = load_unc_model_from_hdf5(nm_path)

fitter.feature_array_flags["empirical_noise_models"] = noise_models

2026-05-01 18:11:35,331 | synference | INFO     | Loaded model from /home/runner/.local/share/Synthesizer/data/synference/BPASS_DenseBasis_v4_final_nsf_0_posterior.pkl.
2026-05-01 18:11:35,332 | synference | INFO     | Device: cpu
2026-05-01 18:11:35,378 | synference | WARNING  | IndexError when trying to set train/test arrays.

Now we need to explain our catalogue to synference. We create a ‘conversion_dict’ dictionary which maps the feature names to columns in the table. The expected column names for flux are the SVO format filter names, e.g. JWST/NIRCam.F070W, or just F070W if it is unambiguous. For flux errors it is the same, but with ‘unc_’ prefixed, e.g. unc_JWST/NIRCam.F070W or unc_F070W. Any other features (e.g. redshift) should be mapped to the column name in the catalogue.

In this case our catalogue is almost in the correct format, so we just need to convert the band names to include the facility and instrument. We do this by getting the band names from the fitted model (fitter.feature_names) and replacing the relevant parts of the strings.

We then create our conversion dictionary to map from e.g. ‘unc_F200W’ to ‘unc_JWST/NIRCam.F200W’.

[4]:

def band_to_instrument(band):
    """Band name to instrument mapping."""
    if band in ["F435W", "F606W", "F775W", "F814W", "F850LP"]:
        return f"HST/ACS_WFC.{band}"
    return f"JWST/NIRCam.{band}"


bands = [
    i.split(".")[-1]
    for i in fitter.feature_names
    if not (i.startswith("unc_") or i.startswith("redshift"))
]

print(bands)

conversion_dict = {band: band_to_instrument(band) for band in bands}
conversion_dict.update({f"unc_{band}": f"unc_{band_to_instrument(band)}" for band in bands})
conversion_dict["redshift"] = "redshift"

conversion_dict

['F435W', 'F606W', 'F775W', 'F814W', 'F850LP', 'F090W', 'F115W', 'F150W', 'F200W', 'F277W', 'F335M', 'F356W', 'F410M', 'F444W']

[4]:

{'F435W': 'HST/ACS_WFC.F435W',
 'F606W': 'HST/ACS_WFC.F606W',
 'F775W': 'HST/ACS_WFC.F775W',
 'F814W': 'HST/ACS_WFC.F814W',
 'F850LP': 'HST/ACS_WFC.F850LP',
 'F090W': 'JWST/NIRCam.F090W',
 'F115W': 'JWST/NIRCam.F115W',
 'F150W': 'JWST/NIRCam.F150W',
 'F200W': 'JWST/NIRCam.F200W',
 'F277W': 'JWST/NIRCam.F277W',
 'F335M': 'JWST/NIRCam.F335M',
 'F356W': 'JWST/NIRCam.F356W',
 'F410M': 'JWST/NIRCam.F410M',
 'F444W': 'JWST/NIRCam.F444W',
 'unc_F435W': 'unc_HST/ACS_WFC.F435W',
 'unc_F606W': 'unc_HST/ACS_WFC.F606W',
 'unc_F775W': 'unc_HST/ACS_WFC.F775W',
 'unc_F814W': 'unc_HST/ACS_WFC.F814W',
 'unc_F850LP': 'unc_HST/ACS_WFC.F850LP',
 'unc_F090W': 'unc_JWST/NIRCam.F090W',
 'unc_F115W': 'unc_JWST/NIRCam.F115W',
 'unc_F150W': 'unc_JWST/NIRCam.F150W',
 'unc_F200W': 'unc_JWST/NIRCam.F200W',
 'unc_F277W': 'unc_JWST/NIRCam.F277W',
 'unc_F335M': 'unc_JWST/NIRCam.F335M',
 'unc_F356W': 'unc_JWST/NIRCam.F356W',
 'unc_F410M': 'unc_JWST/NIRCam.F410M',
 'unc_F444W': 'unc_JWST/NIRCam.F444W',
 'redshift': 'redshift'}

Now we can run the inference. We specify the catalogue table, the conversion dictionary, the input flux units (Jy), and then we have some choices.

We can choose to:

Run predictive checks to remove outliers.
Impute or remove missing data.
Run posterior inference, setting the number of posterior samples to draw.
Recover SEDs for each source.

[5]:

fitter.recreate_simulator_from_library(
    override_library_path=library_path, override_grid_path="test_grid.hdf5"
)

post_tab = fitter.fit_catalogue(
    cat,
    columns_to_feature_names=conversion_dict,
    flux_units=Jy,
    check_out_of_distribution=False,
    recover_SEDs=False,
    missing_data_flag=np.nan,
    num_samples=300,
    append_to_input=False,
)

2026-05-01 18:11:35,427 | synference | INFO     | Overriding internal library name from provided file path.
2026-05-01 18:11:35,701 | synference | WARNING  | Failed to load cosmology from HDF5. Using Planck18 instead.
2026-05-01 18:11:48,163 | synference | INFO     | Updated filters: ['HST/ACS_WFC.F435W', 'HST/ACS_WFC.F606W', 'HST/ACS_WFC.F775W', 'HST/ACS_WFC.F814W', 'HST/ACS_WFC.F850LP', 'JWST/NIRCam.F090W', 'JWST/NIRCam.F115W', 'JWST/NIRCam.F150W', 'JWST/NIRCam.F200W', 'JWST/NIRCam.F277W', 'JWST/NIRCam.F335M', 'JWST/NIRCam.F356W', 'JWST/NIRCam.F410M', 'JWST/NIRCam.F444W']
2026-05-01 18:11:48,166 | synference | INFO     | Simulator recreated from library at /home/runner/.local/share/Synthesizer/data/synference/grid_BPASS_Chab_DenseBasis_SFH_0.01_z_14_logN_2.7_Calzetti_v3_multinode.hdf5.
2026-05-01 18:11:48,167 | synference | INFO     | Auto applying inverse log10 transform for log10_Av.
2026-05-01 18:11:48,167 | synference | INFO     | Auto applying inverse log10 transform for log10_mass_weighted_age.
2026-05-01 18:11:48,168 | synference | INFO     | Auto applying inverse log10 transform for log10_floor_sfr_10.
2026-05-01 18:11:48,169 | synference | INFO     | Adding Av to tau_v transform.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
2026-05-01 18:11:48,179 | synference | INFO     | Removing 54 observations with missing data.

Sampling from posterior: 100%|██████████| 76/76 [00:04<00:00, 17.96it/s]

2026-05-01 18:11:52,415 | synference | INFO     | Obtained posterior samples.

We can look at the output table, which is also an astropy.table.Table.

[6]:

post_tab

[6]:

Table length=99

ID	log_mass_16	log_mass_50	log_mass_84	log10metallicity_16	log10metallicity_50	log10metallicity_84	log10_Av_16	log10_Av_50	log10_Av_84	log_sfr_16	log_sfr_50	log_sfr_84	sfh_quantile_25_16	sfh_quantile_25_50	sfh_quantile_25_84	sfh_quantile_50_16	sfh_quantile_50_50	sfh_quantile_50_84	sfh_quantile_75_16	sfh_quantile_75_50	sfh_quantile_75_84	log10_mass_weighted_age_16	log10_mass_weighted_age_50	log10_mass_weighted_age_84	log10_floor_sfr_10_16	log10_floor_sfr_10_50	log10_floor_sfr_10_84	log_surviving_mass_16	log_surviving_mass_50	log_surviving_mass_84	beta_16	beta_50	beta_84
int64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64	float64
1	9.939584007263184	10.108154296875	10.23990779876709	-3.9572409343719483	-3.8117417097091675	-3.5107469367980957	-0.3456606268882751	-0.16263077408075333	-0.057272515743970874	0.19549835681915287	0.5458541810512543	0.8047231221199036	0.05477115333080292	0.2275593802332878	0.4794836509227753	0.27190587639808655	0.5326043367385864	0.7757581377029419	0.5946298933029175	0.8605380952358246	0.9643951177597045	3.178359317779541	3.3932684659957886	3.52229681968689	0.1732870471477509	0.5576348900794983	0.780247938632965	9.702088012695313	9.848639488220215	9.964367599487305	-1.4193619108200073	-1.2767080664634705	-1.1728471517562866
2	9.154814720153809	9.338487148284912	9.51364387512207	-3.9008369636535645	-3.7287739515304565	-3.4822669982910157	-2.9504598331451417	-2.779890537261963	-2.1422450733184815	-1.1176093578338622	-0.8014236390590668	-0.6317697024345398	0.0678984397649765	0.22708963602781296	0.48523412585258485	0.25436566114425657	0.540723592042923	0.8050130796432495	0.46949163675308225	0.8081181347370148	0.9338459062576294	3.323607006072998	3.525369167327881	3.683005542755127	-1.1443159198760986	-0.8328116238117218	-0.714094090461731	8.901894454956055	9.060781955718994	9.220348281860351	-1.8988433027267455	-1.768225908279419	-1.5384234333038331
3	10.529735603332519	10.657617092132568	10.806660118103027	-3.5730961990356445	-3.0973167419433594	-2.7230493927001955	0.08466074317693711	0.1648702248930931	0.22097436010837554	1.1757683324813843	1.501030683517456	1.6836663055419923	0.0891733142733574	0.2538246810436249	0.5269856905937195	0.3087485980987549	0.6245134174823761	0.826134729385376	0.8128807878494263	0.9549208283424377	0.98389643907547	2.9909888076782227	3.153728485107422	3.319279489517212	1.0783018350601197	1.4581133723258972	1.6440166568756103	10.318914794921875	10.432116031646729	10.565086174011231	-0.6545642995834351	-0.4562164694070816	-0.28958198547363284
4	10.344164009094237	10.49700403213501	10.605947227478028	-2.984836301803589	-2.580446720123291	-2.361547508239746	-1.4467960691452026	-0.6145035922527313	-0.27229827880859375	-1.4194354963302613	-1.0233412384986877	-0.6224310207366943	0.024297930896282197	0.11762548983097076	0.2974019491672516	0.15432520508766173	0.306561678647995	0.5079343128204346	0.30586754441261294	0.4568325877189636	0.5992732715606689	3.5738217926025393	3.6653268337249756	3.73384352684021	-1.486201481819153	-1.1051756143569946	-0.7688006186485291	10.060288772583007	10.202569007873535	10.305483169555664	-0.7077000212669372	0.30610617995262146	0.9208155298233032
5	10.51041374206543	10.603375434875488	10.676143836975097	-3.9544841766357424	-3.7747520208358765	-3.345344514846802	-0.7332632708549499	-0.47819650173187256	-0.3484096038341522	-0.868891875743866	-0.42782431840896606	-0.16015348315238953	0.022929514572024346	0.08696146681904793	0.24596118330955505	0.11459214806556701	0.25826531648635864	0.4611870324611664	0.21432461738586425	0.3811544179916382	0.5998997545242309	3.337169189453125	3.424919366836548	3.477188892364502	-0.8681553077697753	-0.44005441665649414	-0.17796872138977055	10.235732040405274	10.322319030761719	10.388777618408204	-0.4891953158378601	-0.2739820033311844	-0.0966482001543045
6	8.69778305053711	8.904083251953125	9.072546081542969	-3.027278137207031	-2.7472602128982544	-2.614431829452515	-2.5537643527984617	-1.6312091946601868	-0.8682891082763672	-1.0009292697906493	-0.882975846529007	-0.701205825805664	0.05738531693816185	0.18314581364393234	0.4261439597606659	0.23884942531585696	0.48300138115882874	0.7282192230224609	0.4920798134803772	0.7524746954441071	0.9463373303413392	3.5244354724884035	3.708300828933716	3.8192360019683838	-1.01308114528656	-0.9079616665840149	-0.7926965188980103	8.435129814147949	8.621531009674072	8.77653537750244	-2.1178145694732664	-2.0540117025375366	-1.9221822834014892
7	8.270625991821289	8.44380521774292	8.57195743560791	-3.890875835418701	-3.6433757543563843	-3.399082736968994	-2.4719399070739745	-0.9833451509475708	-0.47715588569641115	-0.7157675504684448	-0.5548348426818848	-0.3587524175643921	0.049833663403987885	0.15693114697933197	0.4227946090698242	0.21884294748306274	0.43673717975616455	0.7058078074455261	0.4185807931423187	0.6787207126617432	0.8788844323158265	3.067014570236206	3.2420355081558228	3.337159433364868	-0.7100361251831054	-0.5775145888328552	-0.38637019276618956	8.043687019348145	8.198181629180908	8.314944496154785	-2.23298789024353	-2.147029757499695	-1.945000205039978
8	11.212859420776367	11.323670387268066	11.4198921585083	-3.934935522079468	-3.743446707725525	-3.3498058128356933	0.006739825699478388	0.09191664680838585	0.20431683242321014	0.22936306357383734	0.889330267906189	1.6978036355972288	0.07462345927953722	0.2815987318754196	0.5172363948822022	0.31023181915283204	0.5551360249519348	0.7623483180999756	0.6930764818191528	0.7912870645523071	0.8769993209838867	3.078319721221924	3.2182425260543823	3.3343834018707277	0.2183909630775452	0.8754679560661316	1.6802663707733154	10.974408264160155	11.073630809783936	11.157762069702148	-0.20267665922641753	-0.03926284611225128	0.11406765431165694
9	9.995107879638672	10.10387372970581	10.19513687133789	-3.987198619842529	-3.9449836015701294	-3.7821510314941404	-0.39735090017318725	-0.20022757351398468	-0.017654912620782857	-0.33568749070167536	0.17801903933286667	0.5862455844879151	0.04374103873968125	0.14339467138051987	0.3448799633979797	0.18065268099308013	0.3932361602783203	0.6552660036087036	0.3726856231689453	0.6394424736499786	0.8219149780273437	3.2688940715789796	3.4008145332336426	3.480146913528442	-0.3756799972057342	0.11820689961314201	0.5335055732727051	9.736052894592286	9.83213996887207	9.916845397949219	-1.18173020362854	-1.0558088421821594	-0.8812924289703369
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
90	8.509028205871582	8.62484359741211	8.717207145690917	-3.906067228317261	-3.667941927909851	-3.254606294631958	-2.984197006225586	-2.949751615524292	-2.8277369022369383	-0.02753517262637615	0.04320734553039074	0.1092710319161415	0.07951080083847047	0.2233562096953392	0.4993612802028656	0.4467490839958191	0.7016817927360535	0.8529357814788818	0.822336015701294	0.953859955072403	0.9850706100463867	2.799570569992065	2.963210105895996	3.105194454193115	-0.039396345615386956	0.03511142544448376	0.09528905242681503	8.318460693359375	8.422906398773193	8.499647598266602	-2.3545788764953612	-2.3428192138671875	-2.3179873847961425
91	9.688720397949218	9.750111103057861	9.816822090148925	-3.9633341789245606	-3.896526575088501	-3.7444027996063234	-2.4114700031280516	-1.5445653200149536	-0.6405145835876466	-1.6148673248291014	-0.6706908643245697	0.037952753454446764	0.12486677944660188	0.33409853279590607	0.599639675617218	0.4865436029434204	0.7256273329257965	0.8658796668052673	0.8909614372253418	0.9161343276500702	0.9474739956855774	2.9290439891815185	3.0580238103866577	3.1877545166015624	-1.6564894676208495	-0.6968434751033783	0.02332547470927236	9.471343116760254	9.521464347839355	9.577912788391114	-1.717333436012268	-1.6516600251197815	-1.5686198377609253
92	10.213227310180663	10.352089881896973	10.500415229797364	-2.6412713146209716	-2.2973498106002808	-1.965757040977478	0.22503292441368103	0.28396154940128326	0.33186665892601014	1.2871606636047364	1.4720984101295471	1.6308132791519165	0.048286985456943515	0.17958802729845047	0.3876088356971741	0.19798362374305725	0.4157079756259918	0.6604687452316285	0.385646630525589	0.657511830329895	0.851956820487976	3.149063730239868	3.296972632408142	3.4110130500793456	1.2655787706375121	1.4294313788414001	1.5588575792312622	9.999769096374513	10.11805009841919	10.247216682434082	-0.21020678222179412	0.031191915273666382	0.2637383449077606
93	8.270324668884278	8.41648817062378	8.522477493286132	-3.977976999282837	-3.901645064353943	-3.7265852165222166	-2.1078587150573727	-1.2493868470191956	-0.6279422855377198	-1.7062082052230836	-1.450003445148468	-1.1990451049804687	0.044046077430248264	0.14997181296348572	0.3662202370166778	0.16273850202560425	0.37399089336395264	0.6515532732009888	0.28386874437332155	0.5756015777587891	0.8667307305335998	3.398631019592285	3.5434027910232544	3.63385272026062	-1.693820958137512	-1.4522681832313538	-1.224605040550232	8.0034171295166	8.133708000183105	8.229930152893067	-2.0076331520080566	-1.9071319699287415	-1.7365031480789184
94	8.830239791870117	8.960932731628418	9.080847778320312	-2.684078426361084	-2.217964768409729	-2.119458999633789	-0.3016149890422821	-0.23844961822032928	-0.19485442221164703	0.7584543704986573	0.8984532654285431	0.99396075963974	0.12115693390369416	0.3396892547607422	0.6219125247001648	0.42197991609573365	0.6746797859668732	0.8691137671470642	0.8764693474769593	0.9772589206695557	0.9928487849235534	2.4006847953796386	2.612328052520752	2.7969494915008544	0.819493043422699	0.9105457365512848	0.9940743470191955	8.687822494506836	8.799883365631104	8.910940895080566	-1.7012621879577636	-1.6158613562583923	-1.5505807733535766
95	8.864418296813964	9.073215007781982	9.237068099975586	-3.6159160041809084	-3.1095324754714966	-2.783223934173584	-1.1834826755523682	-0.5715051889419556	-0.323266099691391	-1.6315010070800782	-1.175331711769104	-0.9425984239578247	0.053765920847654344	0.20257627218961716	0.47589844584465024	0.2789442586898804	0.5350314676761627	0.7686481857299805	0.5383857226371765	0.8400853872299194	0.9570275998115539	3.5290346908569337	3.735993504524231	3.8529014015197753	-1.6786646699905396	-1.2243101000785828	-1.0199295139312745	8.594435043334961	8.789204120635986	8.940436515808106	-1.7577277517318726	-1.4093816876411438	-0.9586470198631287
96	10.751171875	10.87028980255127	10.965835037231445	-2.0962821292877196	-1.868673324584961	-1.7033158206939698	-1.0801838874816894	-0.42823173105716705	-0.18155021846294403	-1.307304449081421	-1.0443772673606873	-0.7227890801429748	0.010841950960457326	0.03189085051417351	0.0985772880911827	0.05269346684217453	0.12638313323259354	0.2236345326900482	0.0666769215464592	0.1506529003381729	0.23978349328041076	3.72004563331604	3.7708609104156494	3.796512098312378	-1.5195269012451171	-1.304945468902588	-1.0000944089889527	10.45063003540039	10.564289093017578	10.649251136779785	-0.47832549452781675	0.05412036366760731	0.776248743534088
97	10.827304916381836	10.947047710418701	11.045567741394043	-3.7583420944213866	-3.2069404125213623	-2.7297824668884276	-0.7947887682914734	-0.29402290284633636	-0.12604100346565247	-1.1027163887023925	-0.8602952659130096	-0.6162098073959351	0.010708150751888752	0.03271361626684666	0.08884957075119017	0.049703166484832764	0.1075405552983284	0.19410917341709136	0.061612446159124375	0.13364604115486145	0.230760161280632	3.452608757019043	3.48909330368042	3.513179311752319	-1.1174827241897582	-0.8771901726722717	-0.6248927640914917	10.542222480773926	10.657509803771973	10.749632568359376	1.2977813053131104	1.7334874272346497	2.1066622161865234
98	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan
99	10.622725296020509	10.744435787200928	10.857611045837402	-1.8423300170898438	-1.7279205322265625	-1.6034236097335814	-1.7683616399765014	-1.1998807787895203	-0.9381706190109252	2.384701299667358	2.5009766817092896	2.593200912475586	0.09604063838720323	0.2696179151535034	0.5002139759063721	0.3513221549987793	0.6210521459579468	0.8132014727592468	0.5909414410591126	0.831667572259903	0.969177577495575	2.4139346981048586	2.554788827896118	2.6881879329681397	2.407112464904785	2.49432909488678	2.5698007678985597	10.473794593811036	10.58522653579712	10.691821250915527	-2.101516981124878	-2.0340031385421753	-1.9653551006317138

And we can plot the star-forming main sequence from the fitted catalogue.

[7]:

import matplotlib.pyplot as plt

plt.scatter(post_tab["log_mass_50"], post_tab["log_sfr_50"])
plt.xlabel("Stellar Mass (log M_sun)")
plt.ylabel("SFR (log M_sun/yr)")

[7]:

Text(0, 0.5, 'SFR (log M_sun/yr)')

../_images/posterior_inference_catalogue_fitting_13_1.png

You may notice some of rows and nans. This is because those rows had missing data that we chose not to impute, which can’t be handled by the SBI inference method.

We could set missing_data_mcmc to True to impute those missing values instead.

We can also recover and plot the SEDs, which we’ll do for the first few sources in the catalogue.

[8]:

post_tab, data = fitter.fit_catalogue(
    cat[:4],
    columns_to_feature_names=conversion_dict,
    flux_units=Jy,
    check_out_of_distribution=False,
    recover_SEDs=True,
    plot_SEDs=True,
    missing_data_flag=np.nan,
    num_samples=300,
    append_to_input=False,
)

WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.

Sampling from posterior: 100%|██████████| 4/4 [00:00<00:00, 19.38it/s]

2026-05-01 18:11:52,795 | synference | INFO     | Obtained posterior samples.


Recovering SEDs from posterior samples...:   0%|          | 0/4 [00:00<?, ?it/s]

Starting dense_basis. Failed to load FSPS, only GP-SFH module will be available.
2026-05-01 18:11:52,893 | synference | WARNING  | The following parameters are not used by the simulator: ['log10_Av', 'log_sfr', 'sfh_quantile_25', 'sfh_quantile_50', 'sfh_quantile_75', 'log10_mass_weighted_age', 'log10_floor_sfr_10', 'log_surviving_mass', 'beta', 'Av', 'mass_weighted_age', 'floor_sfr_10']

Recovering SEDs from posterior samples...:   0%|          | 0/4 [00:00<?, ?it/s]

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
Cell In[8], line 1
----> 1 post_tab, data = fitter.fit_catalogue(
      2     cat[:4],
      3     columns_to_feature_names=conversion_dict,
      4     flux_units=Jy,
      5     check_out_of_distribution=False,
      6     recover_SEDs=True,
      7     plot_SEDs=True,
      8     missing_data_flag=np.nan,
      9     num_samples=300,
     10     append_to_input=False,
     11 )

File /opt/hostedtoolcache/Python/3.10.20/x64/lib/python3.10/site-packages/synference/sbi_runner.py:3347, in SBI_Fitter.fit_catalogue(self, observations, columns_to_feature_names, flux_units, missing_data_flag, quantiles, sample_method, sample_kwargs, num_samples, timeout_seconds_per_row, override_transformations, append_to_input, return_feature_array, recover_SEDs, plot_SEDs, check_out_of_distribution, simulator, outlier_methods, missing_data_mcmc, missing_data_mcmc_params, return_full_samples, log_times, use_temp_samples, **kwargs)
   3344     if "." not in param:
   3345         extra_parameters[param] = obs_i[list(self.feature_names).index(param)]
-> 3347 fnu_quantiles, wav, phot_fnu_draws, phot_wav, fig = self.recover_SED(
   3348     X_test=obs_i,
   3349     samples=samples_i,
   3350     num_samples=num_samples,
   3351     sample_method=sample_method,
   3352     sample_kwargs=sample_kwargs,
   3353     plot=plot_SEDs,
   3354     plot_name=f"{self.name}_SED_{pos}",
   3355     simulator=simulator,
   3356     extra_parameters=extra_parameters,
   3357     verbose=False,
   3358 )
   3359 try:
   3360     id = table["ID"][pos]

File /opt/hostedtoolcache/Python/3.10.20/x64/lib/python3.10/site-packages/synference/sbi_runner.py:5833, in SBI_Fitter.recover_SED(self, X_test, samples, num_samples, sample_method, sample_kwargs, posteriors, simulator, prior, plot, marginalized_parameters, extra_parameters, phot_unit, true_parameters, plot_name, plots_dir, sample_color, param_labels, plot_closest_draw_to, plot_sfh, plot_histograms, kde, save_plots, fig, ax, ax_sfh, verbose)
   5831 params.update(extra_parameters)
   5832 params.update(pass_in_observables)
-> 5833 output = simulator(params)
   5834 if i == 0:
   5835     phot_wav = output["photo_wav"]

File /opt/hostedtoolcache/Python/3.10.20/x64/lib/python3.10/site-packages/synference/library.py:6001, in GalaxySimulator.__call__(self, params)
   5999 def __call__(self, params):
   6000     """Call the simulator with parameters to get photometry."""
-> 6001     return self.simulate(params)

File /opt/hostedtoolcache/Python/3.10.20/x64/lib/python3.10/site-packages/synference/library.py:5737, in GalaxySimulator.simulate(self, params)
   5735 if "sfh" in self.output_type:
   5736     stars_sfh = stars.get_sfh()
-> 5737     stars_sfh = stars_sfh / np.diff(10 ** (self.grid.log10age), prepend=0) / yr
   5738     time = 10 ** (self.grid.log10age) * yr
   5739     time = time.to("Myr")

File /opt/hostedtoolcache/Python/3.10.20/x64/lib/python3.10/site-packages/synthesizer/grid.py:336, in Grid.__getattr__(self, name)
    333     return np.log10(self._axes_values[name[5:]])
    335 # If we get here, we don't have the attribute
--> 336 raise AttributeError(
    337     f"'{self.__class__.__name__}' object has no attribute '{name}'"
    338 )

AttributeError: 'Grid' object has no attribute 'log10age'

If you recover the SEDs then a nested dictionary with the SED posteriors are also returned. Initial key is the source index or ID, then within that there are keys for ‘wav’, ‘fnu_quantiles’, ‘wav’, ‘phot_wav’, ‘phot_fnu_draws’ and ‘fig’.

[9]:

data[1].keys()

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[9], line 1
----> 1 data[1].keys()

NameError: name 'data' is not defined

Here are the SED plots:

[10]:

for key in data.keys():
    display(data[key]["fig"])

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[10], line 1
----> 1 for key in data.keys():
      2     display(data[key]["fig"])

NameError: name 'data' is not defined

We can check the feature array used for inference by setting return_feature_array=True.

[11]:

feature_array, mask = post_tab, data = fitter.fit_catalogue(
    cat,
    columns_to_feature_names=conversion_dict,
    flux_units=Jy,
    check_out_of_distribution=False,
    return_feature_array=True,
    missing_data_flag=np.nan,
    num_samples=300,
    append_to_input=False,
)

feature_array

WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
WARNING {'true_flux_units': Jy, 'out_units': 'AB'} arguments will have no effect with this model. Input must be in Jy.
2026-05-01 18:11:54,517 | synference | INFO     | Removing 54 observations with missing data.

[11]:

array([[2.4493113e+01, 2.4163765e+01, 2.3604200e+01, ..., 9.3294436e-04,
        8.9689222e-04, 1.0870000e+00],
       [2.5077990e+01, 2.4549706e+01, 2.3904078e+01, ..., 6.5881172e-03,
        7.1780500e-03, 6.9384003e-01],
       [2.4922672e+01, 2.4493999e+01, 2.3787994e+01, ..., 1.1117980e-03,
        8.8765030e-04, 1.5530000e+00],
       ...,
       [2.6450207e+01, 2.4133791e+01, 2.2739664e+01, ..., 3.6288964e-04,
        2.9280997e-04, 6.6799998e-01],
       [2.7736353e+01, 2.6571774e+01, 2.5916496e+01, ..., 4.7382890e-04,
        3.4990878e-04, 1.6912301e+00],
       [2.3816370e+01, 2.3439030e+01, 2.3062237e+01, ..., 6.9473550e-04,
        6.0235627e-04, 5.7800002e+00]], shape=(76, 29), dtype=float32)

We can also run prior predictive checks to identify outliers, by setting check_out_of_distribution=True. The available list of methods is any of those in PYOD (https://pyod.readthedocs.io/en/latest/). You can set them with the ‘outlier_methods’ argument as a list of strings.