Datasets:

sdrbench
/

hacc

license: other
license_name: sdrbench
license_link: https://sdrbench.github.io/
pretty_name: 'SDRBench: HACC'
tags:
  - scientific-data
  - lossy-compression
  - sdrbench
  - hpc
  - sz3
configs:
  - config_name: files
    data_files: files.jsonl
    default: true

SDRBench — HACC

Cosmology N-body simulation (HACC, ECP ExaSky): particle coordinates xx, yy, zz and velocities vx, vy, vz as separate 1D arrays, single precision, little-endian.

Raw binary arrays, not an Arrow/Parquet dataset. datasets.load_dataset("sdrbench/hacc") returns the table of files (files.jsonl: path, dtype, shape, sha256). Read the arrays with the sdrbench package or hf_hub_download + numpy as shown below.

This repository is an unmodified mirror of the HACC data of SDRBench, the Scientific Data Reduction Benchmark. The originals are hosted by Argonne National Laboratory on Globus; every archive was unpacked and its files uploaded byte-for-byte (sha256 verified) (revision 36891dee00). metadata/ holds the original SDRBench property and template files. Kept in sync automatically by szcompressor/sdrbench.

Data provider: HACC team (ECP ExaSky); https://www.alcf.anl.gov/science/case-studies/opencosmo-community-portal-cosmological-simulations
Size: 150.2 GB in 8 variant(s)
Format: raw little-endian binary, C order (slowest dimension first); dtype and shape per field below

Citation requested by the data provider: Please cite Heitmann et al., Astrophys. J. Suppl. 252 (2021) 2, 19 (https://inspirehep.net/literature/1799008) when using HACC data.

Usage

pip install "sdrbench[sz3]"     # sdrbench + pysz (SZ3)

import numpy as np
import sdrbench
from pysz import sz, szConfig, szErrorBoundMode

ds = sdrbench.dataset("hacc")                # default variant "medium"
print(ds.fields)
x = ds["vx"]                     # numpy array, dtype <f4, shape (280953867,)

# compress with SZ3 (pysz) at a 1e-3 value-range-relative error bound
conf = szConfig()
conf.errorBoundMode = szErrorBoundMode.REL
conf.relErrorBound = 1e-3
compressed, ratio = sz.compress(np.ascontiguousarray(x), conf)
y, _ = sz.decompress(compressed, x.dtype.type, x.shape)
max_err, psnr, nrmse = sz.verify(np.asarray(x), y)
print(f"ratio {ratio:.1f}x, PSNR {psnr:.1f} dB, max error {max_err:.3g}")

Files are fetched on first use (in parallel with ds.download("data/")) and cached, pinned to the revision of the installed sdrbench release; if Hugging Face is unreachable the package falls back to the original archive on Globus. Existing local SDRBench copies can be used with sdrbench.dataset("hacc", root="/path").

Without the package, with huggingface_hub and numpy only:

from huggingface_hub import hf_hub_download
import numpy as np
p = hf_hub_download("sdrbench/hacc", "medium/vx.f32", repo_type="dataset", revision="36891dee003a1aa08bb7da3c3e6b90badcd661e6")
x = np.fromfile(p, dtype="<f4").reshape((280953867,))

Variants and fields

`medium` (default)

sdrbench.dataset("hacc", "medium") — folder medium/ — 8 files, 6.7 GB, from EXASKY-HACC-data-medium-size.tar.gz.

One snapshot with 280,953,867 particles.

Field	dtype	Shape (C order)	File
`vx`	`<f4`	280953867	`medium/vx.f32`
`vy`	`<f4`	280953867	`medium/vy.f32`
`vz`	`<f4`	280953867	`medium/vz.f32`
`xx`	`<f4`	280953867	`medium/xx.f32`
`yy`	`<f4`	280953867	`medium/yy.f32`
`zz`	`<f4`	280953867	`medium/zz.f32`

Other files: medium/EXASKY-HACC-data-medium-size-property.txt, medium/EXASKY-HACC-data-medium-size-template_data.txt

`big`

sdrbench.dataset("hacc", "big") — folder big/ — 8 files, 25.8 GB, from EXASKY-HACC-data-big-size.tar.gz.

One snapshot with 1,073,726,487 particles.

Field	dtype	Shape (C order)	File
`vx`	`<f4`	1073726487	`big/vx.f32`
`vy`	`<f4`	1073726487	`big/vy.f32`
`vz`	`<f4`	1073726487	`big/vz.f32`
`xx`	`<f4`	1073726487	`big/xx.f32`
`yy`	`<f4`	1073726487	`big/yy.f32`
`zz`	`<f4`	1073726487	`big/zz.f32`

Other files: big/EXASKY-HACC-data-big-size-property.txt, big/template_data_big-size.txt

`sbig`

sdrbench.dataset("hacc", "sbig") — folder sbig/ — 3 files, 19.5 GB, from SDRBENCH-EXASKY-HACC-data-sbig-size.tar.gz.

One region (1/768) of time step 499 of a trillion-particle run: coordinates only.

Field	dtype	Shape (C order)	File
`xx`	`<f4`	1621417608	`sbig/27035-xx.f32`
`yy`	`<f4`	1621417608	`sbig/27035-yy.f32`
`zz`	`<f4`	1621417608	`sbig/27035-zz.f32`

`region1`

sdrbench.dataset("hacc", "region1") — folder multi-timesteps/1840/ — 304 files, 19.5 GB, from 1840.tar.gz.

Region 1840: coordinates xx, yy, zz over about 100 time steps (the particle count changes per step) and the particle IDs.