Hugging Face's logo
Activity Feed

AI & ML interests

A community archive making astronomy's measurement-level data — full-sky maps, time streams, event lists, interferometric visibilities, and posterior chains — ML-ready, with provenance back to the canonical archives.

Recent Activity
williamjameshandley  updated a dataset about 6 hours ago
astro-legacy-archive/gwosc-o2-strain
williamjameshandley  updated a dataset about 6 hours ago
astro-legacy-archive/gaia-dr3-source
williamjameshandley  updated a Space about 15 hours ago
astro-legacy-archive/README
View all activity

Organization Card
Community About org cards

Astro Legacy Archive

The Astro Legacy Archive makes astronomy's measurement-level data ready for direct use with standard data and machine-learning tools. The original archives remain authoritative. Each dataset retains the source archive's identifiers, values, shapes, and row order, together with the metadata needed to interpret them.

Why source mirrors matter

Preparing scientific data for a unified machine-learning scheme necessarily selects conventions, equivalences, and priorities from a particular time and purpose. Those transformations are cheap to repeat. Information discarded in the process cannot be recovered, so this archive starts from the canonical release as served instead.

Each mirror preserves the source's identifiers, values, shapes, and row order. It also carries pinned source URLs and hashes, verbatim licence terms, complete structural descriptions, and regeneration provenance. Readers can therefore curate the data reproducibly for their own questions, while any number of harmonised or machine-learning-ready datasets can be derived from—and audited against—the same source mirror. Corrections enter the conversion tools and propagate by regeneration rather than remaining fixed in copied datasets.

Why

Open measurements for cosmology and gravitational-wave inference are scattered across heterogeneous archives, where access can be slow and specialist formats obstruct direct use with standard tools. Here, the source data and their descriptions, units, provenance, and reconstruction instructions travel together. Users can work through load_dataset, streaming, or SQL over HTTP with the source context available in the files.

In place

The organisation currently publishes 71 public datasets. They span radio, microwave, infrared, optical-line and gamma-ray skies, together with detector strain, visibilities, event data and inference products.

Dataset Survey or instrument Band or frequency Kind DOI
caswell-10mhz DRAO Penticton 10.03 MHz map 10.57967/hf/10171
drao-22mhz DRAO 22 MHz T-array 22.25 MHz map 10.57967/hf/10168
ovro-lwa-sky-maps OVRO-LWA 36.528–73.152 MHz map 10.57967/hf/10175
guzman-45mhz Maipú and MU radar surveys 45 MHz map 10.57967/hf/10174
parkes-85mhz Parkes equatorial-strip survey 85 MHz map 10.57967/hf/10172
landecker-wielebinski-150mhz Landecker–Wielebinski survey 150 MHz map 10.57967/hf/10170
eda2-159mhz EDA2 159 MHz map 10.57967/hf/10167
haslam-408mhz Haslam all-sky survey 408 MHz map 10.57967/hf/10163
dwingeloo-820mhz Dwingeloo survey 820 MHz map 10.57967/hf/10169
chipass-1400mhz CHIPASS 1394.5 MHz map 10.57967/hf/10178
reich-1420mhz Stockert and Villa Elisa surveys 1420 MHz map 10.57967/hf/10173
spass-dr1-2.3ghz S-PASS DR1 2303 MHz map 10.57967/hf/10177
rhodes-hartrao-2326mhz Rhodes/HartRAO survey 2326 MHz map 10.57967/hf/10176
wmap-band-maps-9yr WMAP 23, 33, 41, 61, 94 GHz map 10.57967/hf/10165
wmap-ilc-9yr WMAP 23–94 GHz map 10.57967/hf/10164
wmap-masks-9yr WMAP 23–94 GHz analysis masks map 10.57967/hf/10166
wmap-single-year-maps WMAP 23, 33, 41, 61, 94 GHz single-year maps and weight arrays 10.57967/hf/10243
qmap-raw-maps QMAP 26–46 GHz raw sky maps 10.57967/hf/10227
class-released-products CLASS 40 and 90 GHz maps and statistical products 10.57967/hf/10244
acbar-bandpowers ACBAR 150 GHz bandpowers 10.57967/hf/10246
capmap-bandpowers CAPMAP 40 and 90 GHz bandpowers 10.57967/hf/10240
polarbear-released-bandpowers POLARBEAR 150 GHz bandpowers, window functions and covariance matrices 10.57967/hf/10245
bk18-bandpower-covariances BICEP/Keck BK18 multi-frequency bandpower covariance matrices 10.57967/hf/10249
msam-released-products MSAM1 5.6–22.5 cm⁻¹ maps, time streams and statistical products 10.57967/hf/10251
vsa-released-products Very Small Array 26–36 GHz bandpowers, window functions and Hessian matrix 10.57967/hf/10250
planck-frequency-maps-pr3 Planck LFI and HFI, PR3 30–857 GHz map 10.57967/hf/10187
planck-frequency-maps-pr4 Planck LFI and HFI, PR4 NPIPE 30–857 GHz map 10.57967/hf/10188
planck-2018-chains Planck PR3 cosmological-parameter full grid Markov chains 10.57967/hf/10268
cobe-firas-cmb-monopole COBE/FIRAS 2.27–21.33 cm⁻¹ spectrum 10.57967/hf/10181
eht-m87-2017 Event Horizon Telescope 227.0707 and 229.0707 GHz visibilities 10.57967/hf/10183
akari-fir-all-sky-maps AKARI/FIS 65, 90, 140, 160 µm map 10.57967/hf/10185
iris-infrared-maps IRAS/IRIS 12, 25, 60, 100 µm map 10.57967/hf/10186
gaia-dr3-source Gaia DR3 astrometry and photometry source catalogue —
galex-gphoton GALEX FUV and NUV raw photon events and spacecraft state 10.57967/hf/10239
rosat-rass-events ROSAT PSPC-C 0.1–2.4 keV screened events 10.57967/hf/10229
asca-gis2-screened-events ASCA GIS2 0.7–10 keV screened events 10.57967/hf/10242
swift-xrt-cleaned-events Swift XRT 0.3–10 keV cleaned events 10.57967/hf/10233
nicer-xti-cleaned-events NICER XTI 0.2–15 keV cleaned events 10.57967/hf/10235
erosita-dr1-events eROSITA 0.2–10 keV calibrated events 10.57967/hf/10231
hitomi-sxs-cleaned-events Hitomi SXS 0.3–12 keV cleaned events 10.57967/hf/10234
xrism-resolve-cleaned-events XRISM Resolve 0.3–12 keV cleaned events 10.57967/hf/10241
maxi-gsc-events MAXI GSC 2–30 keV medium-telemetry events 10.57967/hf/10238
nustar-fpma-cleaned-events NuSTAR FPMA 3–79 keV cleaned events 10.57967/hf/10237
sas2-photon-events SAS-2 20 MeV–1 GeV photon events 10.57967/hf/10232
fermi-lat-weekly-photons Fermi-LAT 8.439–3,198,155.5 MeV events 10.57967/hf/10182
gw150914-strain LIGO Hanford and Livingston 4,096 Hz sampling strain 10.57967/hf/10180
gwosc-o1-strain LIGO Hanford and Livingston 16,384 Hz sampling strain and data-quality masks 10.57967/hf/10265
gwosc-o2-strain LIGO Hanford, LIGO Livingston, and Virgo 16,384 Hz sampling strain and data-quality masks —
lwa1-lfsm LWA1 Low Frequency Sky Survey 35, 38, 40, 45, 50, 60, 70, 74, 80 MHz maps and error maps 10.57967/hf/10208
arcade2-sky-maps ARCADE 2 3.15–89.8 GHz maps 10.57967/hf/10209
juno-microwave-maps Juno Microwave Radiometer 0.6–21.9 GHz maps and covariance factors 10.57967/hf/10201
quijote-mfi-dr1 QUIJOTE MFI DR1 11, 13, 17, 19 GHz maps 10.57967/hf/10190
cobe-dmr-four-year-sky-maps COBE/DMR 31.5, 53, 90 GHz explicit-pixel maps 10.57967/hf/10216
cobe-dmr-pixelized-differential-data COBE/DMR 31.5, 53, 90 GHz pixelized differential data 10.57967/hf/10217
abs-field-a-maps Atacama B-mode Search 145 GHz explicit-pixel maps 10.57967/hf/10218
abs-bandpowers Atacama B-mode Search 145 GHz bandpowers 10.57967/hf/10210
cbi-released-products Cosmic Background Imager 26–36 GHz bandpowers and matrices 10.57967/hf/10211
tris TRIS 0.6–2.5 GHz CMB-temperature compilation 10.57967/hf/10212
firas-spectral-maps COBE/FIRAS 30–2,880 GHz products explicit-pixel spectral maps 10.57967/hf/10214
dirbe-zsma COBE/DIRBE 1.25–240 µm explicit-pixel maps 10.57967/hf/10213
hi4pi-nhi HI4PI 1420 MHz H I explicit-pixel map 10.57967/hf/10215
hi4pi-low-ivhi HI4PI-derived low/intermediate-velocity H I 1420 MHz H I map 10.57967/hf/10191
ebhis-hi EBHIS 1420 MHz H I map 10.57967/hf/10198
lab-hi-survey LAB H I survey 1420 MHz H I map 10.57967/hf/10197
dickey-lockman-hi Dickey–Lockman H I composite H I column density map 10.57967/hf/10200
dame-co-survey Dame et al. CO survey 115 GHz CO map 10.57967/hf/10189
dame-thaddeus-2022-co CfA CO survey 115 GHz CO map 10.57967/hf/10206
finkbeiner-halpha Finkbeiner H-alpha composite H-alpha maps 10.57967/hf/10196
wham-halpha-map WHAM H-alpha map 10.57967/hf/10199
iras-issa-maps IRAS/ISSA 12, 25, 60, 100 µm maps 10.57967/hf/10204
wise-12um-dust-sample WISE 12 µm dust sample 12 µm grid map 10.57967/hf/10205

Each dataset includes source URLs and SHA-256 pins, licence and credit text, original headers, regeneration commands, and independent verification that its logical source arrays reconstruct bit for bit. Registered dataset DOIs are linked in the table.

The conversion layer now carries production schemas for implicit and explicit pixel tables, grid maps, dense matrices, light curves, visibilities, time streams, statistical products, and event lists (including event-list 1.2.0). Schema metadata travel with each Parquet artifact rather than relying on the dataset card as a machine interface.

Planned

The destination is a source-faithful archive of the measured skies and the native products used to infer from them.

Horizon Programme Destination
Near Complete current bulk releases Grow fermi-lat-weekly-photons from the 145 public weeks w009–w153 toward the complete weekly series through w951.
Mid Complete electromagnetic sky Build a radio-to-gamma-ray collection at or beyond ESASky's coverage directly from canonical measurement-level archive products, with redistribution cleared separately for every source. Include native survey maps and the event or exposure products needed to reconstruct derived intensity maps.
Mid Gravitational waves Extend the O1 release to continuous, per-detector strain and data-quality information for later complete GWOSC observing runs.
Mid Inference products Mirror posterior chains, bandpowers, likelihood inputs, covariance products, and related statistical releases in each collaboration's native names, shapes, ordering, and conventions.
Horizon Other messenger skies Extend the same source-faithful treatment to gravitational-wave event posteriors, neutrino event samples and sky maps, and cosmic-ray arrival directions and exposure maps, so the archive spans electromagnetic, gravitational-wave, neutrino, and cosmic-ray measurements.

What each dataset carries

  • Identifiers, values, shapes, spellings, and row order are exactly those distributed by the source archive.
  • Each column carries a description and, where applicable, a unit. The metadata explain source-specific identifiers and documented codes.
  • Source headers, provenance, and associated structures travel with the measurements and support reconstruction.
  • The logical arrays in the original source files reconstruct bit for bit.
  • Versioned conversion software and pinned source hashes reproduce each published file.

Scope

The archive focuses on measurement- and instrument-level products: full-sky maps, time streams, event lists, visibilities, and statistical products. For object-level collections of images, spectra, and light curves, see the Multimodal Universe. Other organisations are welcome to mirror these datasets.

Errata, corrections, and requests

For an erratum or proposed correction to a particular dataset, open a Discussion or Pull Request in that dataset's Community tab. For dataset requests and questions about the archive as a whole, use the README space's Community tab or the Community tab of any dataset. Posting requires a free Hugging Face account. By default, Hub notifications alert organisation members to new activity. Confirmed corrections are made in the source conversion tools and published by rebuilding the dataset.

Credit

Dataset cards identify the instruments, publications, and source archives to credit and cite. This work depends on decades of stewardship by observatories, collaborations, and archive teams, including NASA LAMBDA/HEASARC/IRSA, ESA, CDS, GWOSC, and their colleagues.

Collections 8

View 8 collections

models 0

None public yet