Official SPT data for the differentiable CMB likelihood framework candl.
To install the SPT candl data library, simply navigate to where you would like to store the data and then run:
git clone https://github.com/SouthPoleTelescope/spt_candl_data.git cd spt_candl_data pip install .
This will download the relevant data files. The installation gives you access to handy short cuts that make it easier to initialise the likelihoods.
Note that you also need to install candl in order to run the likelihoods, run pip install candl or see the repository for more detailed instructions.
To check that everything is working, you can test the supplied likelihoods (provided you have already installed candl) via:
import spt_candl_data spt_candl_data.run_all_tests()
| Name | Key Shortcuts | References | candl version used for publication |
|---|---|---|---|
| SPT-3G D1 T&E | spt_candl_data.SPT3G_D1_TnE, spt_candl_data.SPT3G_D1_TnE_lite |
Quan et al. 2025 (in prep.) |
2.0.3 |
| SPT-3G D1 BB | spt_candl_data.SPT3G_D1_BB |
Zebrowski et al. 2025 | 2.1.0 |
You can also get a detailed summary of the variants of all likelihoods available, by running the following python code:
import spt_candl_data spt_candl_data.print_all_shortcuts()
For subsets of the SPT-3G D1 T&E likelihood, such as fitting only EE spectra, use the corresponding variant of the likelihood. Avoid using the candl keyword data_selection, unless for restricting the angular scales used.
We supply files to help you use SPT data with common cosmological samplers as well as tutorials on how to interact with the data and perform common analysis tasks. You can find more help and tutorials in the candl documentation.
tutorial_notebooks/SPT3G_D1_TnE_tutorial.ipynb: this notebook uses the SPT-3G D1 T&E likelihoods and is split into two parts. The first one shows you how to initialize the multifrequency likelihood, evaluate it, visualize the data, and helps you understand the data model. The second part uses the SPTlite likelihood and leverages the differentiability of candl. This part shows how to translate biases from the band-power level to the parameter-level and how to perform gradient-based minimization and sampling.
tutorial_notebooks/SPT3G_D1_BB_tutorial.ipynb: this notebook uses the SPT-3G D1 BB likelihood. It shows you how to interact with the data, run chains using Cobaya, and demonstrates the differentiability of the likelihood.
You can find template Cobaya .yaml files to help you launch chains as well as ΛCDM proposal matrices in the cobaya/ folder
Montepython likelihood folders you can copy-paste into your MontePython installation are in the montepython/ folder.
For each likelihood, there is a template .param file and a ΛCDM proposal matrix.
Note that the MontePython likelihoods by default do not remove the candl internal priors. For more information, see the candl documentation.
You can find template CosmoSIS .ini files to help you launch chains in the cosmosis/ folder.
Please cite:
- the paper(s) of the relevant likelihood(s) you use (see the data summary table above for quick reference) and
- the candl release paper.
