Datasets:

astroshawn
/

SpecCLIP-TestData-Small

license: cc-by-4.0
task_categories:
  - feature-extraction
tags:
  - astronomy
  - spectroscopy
  - gaia
  - lamost
  - stellar-spectra
pretty_name: SpecCLIP Test Data
size_categories:
  - 1K<n<10K

SpecCLIP Test Data

Test data for SpecCLIP - Cross-modal stellar spectral analysis.

Dataset Description

This dataset contains paired Gaia XP and LAMOST LRS spectra for testing cross-modal retrieval and embedding analysis.

Files

gaia_lamost_test_only.h5: HDF5 file containing paired spectra
- test/source_ids: Gaia source identifiers
- test/gaia_spectra: Gaia XP spectra (N, 343)
- test/lamost_spectra: LAMOST LRS spectra (N, 1462)

Usage

Download

from huggingface_hub import hf_hub_download

# Download test data
test_data_path = hf_hub_download(
    repo_id="astroshawn/SpecCLIP-TestData-Small",
    filename="gaia_lamost_test_only.h5",
    repo_type="dataset",
    local_dir="./test_data"
)

Or using CLI:

huggingface-cli download astroshawn/SpecCLIP-TestData-Small --local-dir ./test_data --repo-type dataset

Load Data

import h5py
import numpy as np

with h5py.File(test_data_path, 'r') as f:
    source_ids = np.array(f['test/source_ids'][:])
    gaia_spectra = np.array(f['test/gaia_spectra'][:])
    lamost_spectra = np.array(f['test/lamost_spectra'][:])

print(f"Number of test samples: {len(source_ids)}")
print(f"Gaia XP spectra shape: {gaia_spectra.shape}")
print(f"LAMOST LRS spectra shape: {lamost_spectra.shape}")

Use with SpecCLIP

This test data is used for:

Building embedding databases for spectral retrieval
Cross-modal spectrum prediction
In-modal and cross-modal similarity search

See the SpecCLIP model repository for usage examples.

Citation

If you use this data, please cite the SpecCLIP paper.