variant stringclasses 2
values | field stringclasses 2
values | var stringclasses 2
values | step null | path stringclasses 6
values | dtype stringclasses 1
value | shape listlengths 3 3 ⌀ | file_shape listlengths 3 3 ⌀ | component null | transpose null | bytes int64 2.15k 16.8B | sha256 stringclasses 6
values |
|---|---|---|---|---|---|---|---|---|---|---|---|
small | null | null | null | basilisk/metadata_1024x1024x1501Re10k.txt | null | null | null | null | null | 2,151 | 14ca2c0901f0bda562228e1bd0af174d1ab5c96749f84b0705d7570cc1060afc |
small | u | u | null | basilisk/u_1024x1024x1501Re10k.f32 | <f4 | [
1501,
1024,
1024
] | [
1501,
1024,
1024
] | null | null | 6,295,650,304 | 052412b665a336707c6ea22cfae8a6ff00dddefa01400f37b9d2fd1a97b36e07 |
small | v | v | null | basilisk/v_1024x1024x1501Re10k.f32 | <f4 | [
1501,
1024,
1024
] | [
1501,
1024,
1024
] | null | null | 6,295,650,304 | 1699cd12f48bb167141ddd11cf2819bb581f0562426264f106ac4745de4248ad |
large | null | null | null | basilisk/metadata_2048x2048x1001Re10k.txt | null | null | null | null | null | 2,156 | 1abee5d1a3acd59c2377c1ebe34952f286a0366ae172d747fa6978062e10dc4e |
large | u | u | null | basilisk/u_2048x2048x1001Re10k.f32 | <f4 | [
1001,
2048,
2048
] | [
1001,
2048,
2048
] | null | null | 16,793,993,216 | 8bd2a150a12e49a2c487f86130c593a70901d456ec4290741f303926694e85b5 |
large | v | v | null | basilisk/v_2048x2048x1001Re10k.f32 | <f4 | [
1001,
2048,
2048
] | [
1001,
2048,
2048
] | null | null | 16,793,993,216 | fd50be4192629a686690fc662e657699a9c9848c825ac9f406a99e38a2c02d16 |
SDRBench — Basilisk-Turbulence
2D decaying turbulence at Re = 10,000 simulated with Basilisk (http://basilisk.fr/): velocity components u, v over time, single precision, little-endian. The first time step is an all-zero placeholder.
Raw binary arrays, not an Arrow/Parquet dataset.
datasets.load_dataset("sdrbench/basilisk-turbulence")returns the table of files (files.jsonl: path, dtype, shape, sha256). Read the arrays with thesdrbenchpackage orhf_hub_download+ numpy as shown below.
This repository is an unmodified mirror of the Basilisk-Turbulence 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 9b98ad8cb9). metadata/ holds
the original SDRBench property and template files. Kept in sync automatically by
szcompressor/sdrbench.
- Data provider: M. Xia, Y. Li, P. Jiao, B. Wang, X. Liang and H. Guo, "Time-Varying Vector Field Compression with Preserved Critical Point Trajectories," IEEE ICDE 2026, pp. 2235-2249, doi:10.1109/ICDE65706.2026.00168
- Size: 46.2 GB in 2 variant(s)
- Format: raw little-endian binary, C order (slowest dimension first); dtype and shape per field below
Usage
pip install "sdrbench[sz3]" # sdrbench + pysz (SZ3)
import numpy as np
import sdrbench
from pysz import sz, szConfig, szErrorBoundMode
ds = sdrbench.dataset("basilisk-turbulence") # default variant "small"
print(ds.fields)
x = ds["u"] # numpy array, dtype <f4, shape (1501, 1024, 1024)
# 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("basilisk-turbulence", 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/basilisk-turbulence", "basilisk/u_1024x1024x1501Re10k.f32", repo_type="dataset", revision="9b98ad8cb9a44ff91e46ebcd85ac74ffd3d4373a")
x = np.fromfile(p, dtype="<f4").reshape((1501, 1024, 1024))
Variants and fields
small (default)
sdrbench.dataset("basilisk-turbulence", "small") — folder basilisk/ — 3 files, 12.6 GB, from SDRBENCH-Basilisk-Turbulence.tar.gz.
1024 x 1024 grid, 1501 time steps: shape (time, y, x).
| Field | dtype | Shape (C order) | File |
|---|---|---|---|
u |
<f4 |
1501 x 1024 x 1024 | basilisk/u_1024x1024x1501Re10k.f32 |
v |
<f4 |
1501 x 1024 x 1024 | basilisk/v_1024x1024x1501Re10k.f32 |
Other files: basilisk/metadata_1024x1024x1501Re10k.txt
large
sdrbench.dataset("basilisk-turbulence", "large") — folder basilisk/ — 3 files, 33.6 GB, from SDRBENCH-Basilisk-Turbulence.tar.gz.
2048 x 2048 grid, 1001 time steps: shape (time, y, x).
| Field | dtype | Shape (C order) | File |
|---|---|---|---|
u |
<f4 |
1001 x 2048 x 2048 | basilisk/u_2048x2048x1001Re10k.f32 |
v |
<f4 |
1001 x 2048 x 2048 | basilisk/v_2048x2048x1001Re10k.f32 |
Other files: basilisk/metadata_2048x2048x1001Re10k.txt
Citation
Please cite SDRBench; sdrbench cite basilisk-turbulence prints it:
@inproceedings{zhao2020sdrbench,
title = {{SDRBench}: Scientific Data Reduction Benchmark for Lossy Compressors},
author = {Zhao, Kai and Di, Sheng and Liang, Xin and Li, Sihuan and Tao, Dingwen and Chen, Zizhong and Cappello, Franck},
booktitle = {2020 IEEE International Conference on Big Data (Big Data)},
pages = {2716--2724},
year = {2020}
}
License and terms
The data are distributed by the SDRBench team for research use under the terms of the original data providers listed above (see the SDRBench website); please acknowledge them as requested. Rights remain with the data providers.
- Downloads last month
- 43