torchsky
Sky coordinates, spherical geometry, and mapmaking as PyTorch tensor operations. They run on CPU, CUDA, and MPS, support autograd, and scale to survey-sized inputs.
torchsky covers FITS-WCS, HEALPix, spherical harmonic transforms, catalog crossmatch, image coadds, spectral cubes, and observation operators. FITS reading and writing is handled by torchfits, which torchsky installs as a dependency.
Install
Stable releases, from 0.5.0, are on PyPI: pip install torchsky. Release
candidates are only GitHub pre-releases
and are never uploaded to PyPI. Building from source needs a C++ compiler and
CMake for the native extension:
git clone https://github.com/astroai/torchsky.git
cd torchsky
pip install .
torchsky supports Python 3.12 to 3.14 with PyTorch 2.14. The installation guide covers CUDA, Pixi, and troubleshooting.
Quick start
import torch
import torchsky.sphere as sphere
from torchsky.map import Observation, coadd, plan
from torchsky.wcs import WCS
header = {"NAXIS": 2, "NAXIS1": 64, "NAXIS2": 64,
"CTYPE1": "RA---TAN", "CTYPE2": "DEC--TAN", "CRPIX1": 32.5, "CRPIX2": 32.5,
"CRVAL1": 150.0, "CRVAL2": 2.0, "CDELT1": -2.8e-4, "CDELT2": 2.8e-4}
ra, dec = WCS(header).pixel_to_world(torch.tensor([0.0, 63.0]), torch.tensor([0.0, 63.0]))
pix = sphere.ang2pix(1024, ra, dec, lonlat=True, nest=True)
obs = [Observation(data=torch.randn(64, 64), wcs={**header, "CRVAL1": c}) for c in (150.0, 150.01)]
result = coadd(obs, plan(obs, projection="TAN"), kernel="bilinear")
For real files, read them with torchfits and pass the header on:
data, header = torchfits.read("image.fits", return_header=True).
What's in the box
| Namespace | Contents |
|---|---|
torchsky.wcs |
FITS-WCS projections, SIP/TPV distortions, affine fitting |
torchsky.sphere |
HEALPix, MOC, sparse maps, polygons, spin SHT |
torchsky.coords |
Frame conversion, observed alt/az, proper motion |
torchsky.catalogs |
Crossmatch and spherical partitioning |
torchsky.map |
Coadds, spectral cubes, observation operators, solvers |
torchsky.spectral and friends |
Spectra, photometry, extinction, cosmology |
torchsky.scene, torchsky.field |
Forward models and continuous sky fields (experimental) |
Performance
On 135 JWST/NIRCam exposures (566 Mpix) onto a 512 x 512 grid, torchsky's Lanczos-3 coadd takes 14.3 s on 8 CPU threads and 5.2 s on a GPU slice, against 30.1 s for SWarp. Its drizzle matches STScI drizzle to better than 2e-6 sigma. A 280-CCD UNIONS tile coadd fell from 2292 s to 711 s on CPU. Hardware, versions, and evidence are on the performance highlights page.
Documentation
- Quick start and examples
- User guide and API reference (rendered at astroai.github.io/torchsky)
- Coming from Astropy, healpy, or reproject
- Changelog
Contributing
Development uses pixi:
git clone https://github.com/astroai/torchsky.git && cd torchsky
pixi install --locked
pixi run dev # build the native extension in place
pixi run preflight-push # lint + compile check
pixi run test # unit suite
Read CONTRIBUTING.md before opening a PR. Issues labelled good first issue are a good place to start.
Citation
If torchsky helps your research, please cite it using CITATION.cff.
License
1 maintainer
Metadata
1 maintainer
Release files for torchsky 0.5.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| torchsky-0.5.0.tar.gz | 2.1 MB | With provenance | Details for torchsky-0.5.0.tar.gz |
Built distributions (wheels)
Total release size: 13.3 MB
Release files / torchsky-0.5.0.tar.gz
| Download URL | torchsky-0.5.0.tar.gz |
|---|---|
| Size | 2.1 MB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
171bf6df21d3cdb7d74f22748c9430ec3a869544e703894fa86f1444b8e3d450
|
|
BLAKE2b-256 checksum How to use checksums |
2072eea5f8819ca0702e0bc0b13e16a7ebae4289ce139764e25c136e2ba45ba8
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 3, 2026.
Transparency logRelease files / torchsky-0.5.0-cp314-cp314-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
| Download URL | torchsky-0.5.0-cp314-cp314-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl |
|---|---|
| Size | 1.9 MB |
| Tags | CPython 3.14 Linux glibc 2.27+ x86-64 Linux glibc 2.28+ x86-64 |
|
SHA-256 checksum How to use checksums |
8214f5fabade62ce68dd2fde2e0b28205cd56ef0a8f7fae1f450be7345065d8f
|
|
BLAKE2b-256 checksum How to use checksums |
d3cb5f6f1398ab8aadf5958412e416894ab190621aea6aeaa34aed9f9a49921c
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 3, 2026.
Transparency logRelease files / torchsky-0.5.0-cp314-cp314-macosx_14_0_arm64.whl
| Download URL | torchsky-0.5.0-cp314-cp314-macosx_14_0_arm64.whl |
|---|---|
| Size | 1.8 MB |
| Tags | CPython 3.14 macOS 14.0+ ARM64 |
|
SHA-256 checksum How to use checksums |
2bdc99a7f9e4b1a43cf6c2d9c848e93e2a9f32b5a871ecd1740109752e4a193c
|
|
BLAKE2b-256 checksum How to use checksums |
8143816a4f5ee330b83c125a21c2cfd1238cef0036f1be09d4c098d0414aac6f
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 3, 2026.
Transparency logRelease files / torchsky-0.5.0-cp313-cp313-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
| Download URL | torchsky-0.5.0-cp313-cp313-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl |
|---|---|
| Size | 1.9 MB |
| Tags | CPython 3.13 Linux glibc 2.27+ x86-64 Linux glibc 2.28+ x86-64 |
|
SHA-256 checksum How to use checksums |
3b2f51ab518380d6d7f25913d43138416f5e7ae003298fd72d4c404e5e81dac6
|
|
BLAKE2b-256 checksum How to use checksums |
351960d7110554d9308c765dbaad3e8d7c73f5cbae3b5c67bc1a03cb2f1fa6fa
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 3, 2026.
Transparency logRelease files / torchsky-0.5.0-cp313-cp313-macosx_14_0_arm64.whl
| Download URL | torchsky-0.5.0-cp313-cp313-macosx_14_0_arm64.whl |
|---|---|
| Size | 1.8 MB |
| Tags | CPython 3.13 macOS 14.0+ ARM64 |
|
SHA-256 checksum How to use checksums |
ecdd59d870bcb8706b3b36d23ea306be97120659226e2f9b4fdc2969aa68793a
|
|
BLAKE2b-256 checksum How to use checksums |
6abe16801a65ba7167afbd4062e7ebb360cb94d5ef6ee67d73251de03cbbe536
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 3, 2026.
Transparency logRelease files / torchsky-0.5.0-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
| Download URL | torchsky-0.5.0-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl |
|---|---|
| Size | 1.9 MB |
| Tags | CPython 3.12 Linux glibc 2.27+ x86-64 Linux glibc 2.28+ x86-64 |
|
SHA-256 checksum How to use checksums |
20e0bc2acc55940f685cbbffee24c1f8e017fa1cd124c2dfe2331901412e7522
|
|
BLAKE2b-256 checksum How to use checksums |
728cf6b4bcf9248b78cee38a98ed0b75613225d465901a430b94aa863fd9d9e7
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 3, 2026.
Transparency logRelease files / torchsky-0.5.0-cp312-cp312-macosx_14_0_arm64.whl
| Download URL | torchsky-0.5.0-cp312-cp312-macosx_14_0_arm64.whl |
|---|---|
| Size | 1.8 MB |
| Tags | CPython 3.12 macOS 14.0+ ARM64 |
|
SHA-256 checksum How to use checksums |
114b9f1a6585ff8c5f1758dfd54929390c7994f3f99bda4325bc86649a719ee3
|
|
BLAKE2b-256 checksum How to use checksums |
3edc038075f64c115a0f53ca0c3d9a6175c226092ce719c5d30c0cb700cd1be5
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 3, 2026.
Transparency logSecurity for torchsky 0.5.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
Full provenance
All 7 files in this release have been uploaded with provenance metadata.
Attestation for torchsky-0.5.0.tar.gz
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 3, 2026.
Transparency logA note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.
