Language: English | 简体中文
TransFit is a Python package for forward modeling and fitting astronomical transient light curves. It provides a compact interface for bolometric and multi-band data, with built-in nickel, magnetar, magnetar-plus-nickel, and CSM interaction models.
- Physical light-curve models with bolometric luminosity, effective temperature, and photospheric radius outputs.
- Multi-band photometry in flux or magnitude space, including filter, extinction, and SED handling.
- Bayesian fitting through a consistent result object, with
emceeinstalled by default and optionalzeusanddynestybackends.
python -m pip install transfitFor local development:
git clone <your-repo-url>
cd TransFit
python -m pip install -e ".[plot,examples]"Install optional sampler backends with:
python -m pip install "transfit[all-samplers]"Forward bolometric light curve
import matplotlib.pyplot as plt
import transfit as tf
params = {
"M_ej": 3.0,
"v_ej": 1.0,
"E_Th_in": 1.5,
"M_ni": 0.08,
"R_0": 120.0,
"f_ni": 0.2,
"kappa": 0.12,
"kappa_gamma": 0.03,
"T_floor": 4500.0,
}
lc = tf.lightcurve_bol(
model="nickel",
params=params,
z=0.001728,
t_max_days=120.0,
)
plt.plot(lc.t_days, lc.Lbol)
plt.yscale("log")
plt.xlabel("Observer-frame time (days)")
plt.ylabel("Bolometric luminosity (erg s$^{-1}$)")
plt.show()
Forward multi-band light curve
import matplotlib.pyplot as plt
import transfit as tf
params = {
"M_ej": 3.0,
"v_ej": 1.0,
"E_Th_in": 1.5,
"M_ni": 0.08,
"R_0": 120.0,
"f_ni": 0.2,
"kappa": 0.12,
"kappa_gamma": 0.03,
"T_floor": 4500.0,
}
filters = {
"B": "johnson_cousins.B",
"V": "johnson_cousins.V",
"R": "johnson_cousins.R",
"I": "johnson_cousins.I",
}
lc = tf.lightcurve_multiband(
model="nickel",
params=params,
z=0.001728,
filters=filters,
bands=["B", "V", "R", "I"],
y_kind="mag",
mag_system="vega",
t_max_days=120.0,
)
for band in lc.bands:
plt.plot(lc.t_days, lc.y[band], label=band)
plt.gca().invert_yaxis()
plt.xlabel("Observer-frame time (days)")
plt.ylabel("Vega magnitude")
plt.legend()
plt.show()filters maps the band labels in your data to filter definitions. Built-in
filters use string IDs. Custom mono filters should use an effective wavelength:
filters = {
"g": {"lambda_eff_A": 4770.0},
"r": {"lambda_eff_nm": 623.1},
}For custom Vega magnitudes, also provide a Vega zero point:
filters = {
"B": {"lambda_eff_A": 4400.0, "vega_zero_point_jy": 4260.0},
}
Fit a bolometric light curve
import numpy as np
import transfit as tf
arr = np.loadtxt("examples/data/sn1993j_lbol.txt")
data = tf.BolometricData(
t_days=arr[:, 0] - arr[:, 0].min(),
y=arr[:, 1],
yerr=arr[:, 2],
)
res = tf.fit_bol(
data=data,
model="nickel",
z=0.001728,
priors={
"M_ej": (0.5, 8.0),
"v_ej": (0.2, 3.0),
"E_Th_in": (0.05, 8.0),
"M_ni": ("log10", -3.0, -0.2),
"R_0": (10.0, 400.0),
"t_shift": (0.0, 20.0),
},
fixed={
"f_ni": 0.2,
"kappa": 0.12,
"kappa_gamma": 0.03,
},
sampler_kwargs={"nwalkers": 32, "nsteps": 5000, "burnin": 1000, "thin": 10},
)
print(res.best_params_raw)
tf.save(res, "mcmc_out/sn1993j_bol_nickel.npz")Fit a multi-band light curve
import numpy as np
import transfit as tf
raw = np.genfromtxt(
"examples/data/sn2007gr.csv",
delimiter=",",
names=True,
dtype=float,
encoding="utf-8",
)
bands, t_days, y, yerr = [], [], [], []
t0 = np.nanmin(raw["Phase"])
columns = {
"B": ("Bmag", "e_Bmag"),
"V": ("Vmag", "e_Vmag"),
"R": ("Rmag", "e_Rmag"),
"I": ("Imag", "e_Imag"),
}
for band, (mag_col, err_col) in columns.items():
good = (
np.isfinite(raw["Phase"])
& np.isfinite(raw[mag_col])
& np.isfinite(raw[err_col])
& (raw[err_col] > 0)
)
t_days.extend((raw["Phase"][good] - t0).tolist())
y.extend(raw[mag_col][good].tolist())
yerr.extend(raw[err_col][good].tolist())
bands.extend([band] * int(np.sum(good)))
data = tf.MultiBandData(
t_days=np.asarray(t_days, float),
band=np.asarray(bands, dtype=object),
y=np.asarray(y, float),
yerr=np.asarray(yerr, float),
)
filters = {
"B": "johnson_cousins.B",
"V": "johnson_cousins.V",
"R": "johnson_cousins.R",
"I": "johnson_cousins.I",
}
res = tf.fit_multiband(
data=data,
model="nickel",
z=0.001728,
filters=filters,
y_kind="mag",
mag_system="vega",
priors={
"M_ej": (0.5, 8.0),
"v_ej": (0.2, 3.0),
"E_Th_in": (0.05, 8.0),
"M_ni": ("log10", -3.0, -0.2),
"R_0": (10.0, 400.0),
"t_shift": (0.0, 20.0),
},
fixed={
"f_ni": 0.2,
"kappa": 0.12,
"kappa_gamma": 0.03,
"T_floor": 4500.0,
},
sampler_kwargs={"nwalkers": 32, "nsteps": 5000, "burnin": 1000, "thin": 10},
)
print(res.best_params_raw)
tf.save(res, "mcmc_out/sn2007gr_multiband_nickel.npz")Data containers
tf.BolometricData(t_days, y, yerr, mask=None)
tf.MultiBandData(t_days, band, y, yerr, mask=None)Model inspection
tf.model_param_names("nickel")
tf.param_template("csm")Canonical model names are nickel, magnetar, magnetar_ni, and csm.
Forward and prediction
tf.lightcurve_bol(model=..., params=..., z=..., t_max_days=...)
tf.lightcurve_multiband(
model=...,
params=...,
z=...,
filters=...,
bands=...,
y_kind="mag",
)
tf.predict_bol(model=..., params=..., z=..., t_days=...)
tf.predict_multiband(
model=...,
params=...,
z=...,
filters=...,
t_days=...,
band=...,
)Fitting
tf.fit_bol(
data=...,
model=...,
z=...,
priors=...,
fixed=...,
sampler="emcee",
sampler_kwargs=None,
model_kwargs=None,
)
tf.fit_multiband(
data=...,
model=...,
z=...,
filters=...,
y_kind="mag",
priors=...,
fixed=...,
sed=None,
sampler="emcee",
sampler_kwargs=None,
model_kwargs=None,
)Results, plotting, and I/O
res.best_params
res.best_params_raw
res.median_params
res.best_fit
res.best_index
res.best_log_prob
res.best_sample
res.samples
res.log_prob
res.meta
tf.plot.fit_bol(res, data=data)
tf.plot.fit_multiband(res, data=data)
tf.plot.corner(res)
path = tf.save(res, path="mcmc_out/result.npz")
loaded = tf.load(path)Full details are available in API and parameter reference.
For questions about this project, please contact:
- Liangduan Liu (liuld@ccnu.edu.cn)
- Yuhao Zhang (zhangyh2001@foxmail.com)
- GuangLei Wu (wuguanglei@mails.ccnu.edu.cn)
- Jingyao Li (lijy@mails.ccnu.edu.cn)
If TransFit is helpful for your work, please consider giving the repository a star. This helps other researchers discover the project.
If you use TransFit in research, please cite the TransFit paper:
@ARTICLE{2025ApJ...992...20L,
author = {{Liu}, Liang-Duan and {Zhang}, Yu-Hao and {Yu}, Yun-Wei and {Du}, Ze-Xin and {Li}, Jing-Yao and {Wu}, Guang-Lei and {Dai}, Zi-Gao},
title = "{TransFit: An Efficient Framework for Transient Light-curve Fitting with Time-dependent Radiative Diffusion}",
journal = {\apj},
keywords = {Supernovae, Radiative transfer, Core-collapse supernovae, Time domain astronomy, 1668, 1335, 304, 2109, High Energy Astrophysical Phenomena, Instrumentation and Methods for Astrophysics},
year = 2025,
month = oct,
volume = {992},
number = {1},
eid = {20},
pages = {20},
doi = {10.3847/1538-4357/adfed6},
archivePrefix = {arXiv},
eprint = {2505.13825},
primaryClass = {astro-ph.HE},
adsurl = {https://ui.adsabs.harvard.edu/abs/2025ApJ...992...20L},
adsnote = {Provided by the SAO/NASA Astrophysics Data System}
}For the csm model, also cite:
@ARTICLE{2026ApJ...999..186Z,
author = {{Zhang}, Yu-Hao and {Liu}, Liang-Duan and {Du}, Ze-Xin and {Wu}, Guang-Lei and {Li}, Jing-Yao and {Yu}, Yun-Wei},
title = "{TransFit-CSM: A Fast, Physically Consistent Framework for Interaction-powered Transients}",
journal = {\apj},
keywords = {Core-collapse supernovae, Supernovae, Circumstellar matter, Stellar mass loss, 304, 1668, 241, 1613, High Energy Astrophysical Phenomena},
year = 2026,
month = mar,
volume = {999},
number = {2},
eid = {186},
pages = {186},
doi = {10.3847/1538-4357/ae434a},
archivePrefix = {arXiv},
eprint = {2511.13265},
primaryClass = {astro-ph.HE},
adsurl = {https://ui.adsabs.harvard.edu/abs/2026ApJ...999..186Z},
adsnote = {Provided by the SAO/NASA Astrophysics Data System}
}- Ni et al., Mapping the Dense Circumstellar Environments of SNe Ibn, SNe Icn, and Fast Blue Optical Transients, arXiv e-print (2026), arXiv:2607.00453.
- Yuan et al., Thermal X-rays breaking out from pre-explosion ejecta of a dying massive star, arXiv e-print (2026), arXiv:2606.10014.
- Liu et al., SN 2024igg: A Super-Chandrasekhar/03fg-like SN exhibiting C II-dominated spectra after explosion, submitted to A&A (2026), arXiv:2602.03427.
Parts of the code and documentation in this project were generated with assistance from OpenAI Codex.