Skip to content

DATA_RELEASES.md

Latest commit

 

History

History
77 lines (59 loc) · 2.34 KB

DATA_RELEASES.md

File metadata and controls

77 lines (59 loc) · 2.34 KB

ELFNet Data Releases

ELFNet stores large datasets as compressed archives tracked with Git LFS. Regular Git history contains code, small examples, documentation, checksums, and LFS pointers.

Datasets

Dataset Contents Path
pressure-triplets-326k-v1 326,009 pressure SAD/ELF/symmetry triplets, 978,027 .npy files release/pressure-triplets-326k-v1/
dft-reference-elfs-75k-v1 75,000 DFT reference ELFCAR files release/dft-reference-elfs-75k-v1/

Download With Git LFS

Install Git LFS before cloning, or run git lfs pull after cloning.

git lfs install
git clone git@github.com:Austin243/ELFNet.git
cd ELFNet
git lfs pull --include="release/**"

To fetch only one dataset:

git lfs pull --include="release/pressure-triplets-326k-v1/**"

Extract DFT Reference Archive

dft-reference-elfs-75k-v1 uses one split archive. Verify the part checksums, reassemble the archive, then extract it:

cd release/dft-reference-elfs-75k-v1
sha256sum -c SHA256SUMS
cat dft-reference-elfs-75k-v1.tar.zst.part-* > dft-reference-elfs-75k-v1.tar.zst
sha256sum -c SHA256SUMS.full
tar --use-compress-program=unzstd -xf dft-reference-elfs-75k-v1.tar.zst

Extract Pressure 326k Shards

pressure-triplets-326k-v1 is split into 32 shard archives. Shard 006 is itself split into three parts to keep every LFS object below 2 GiB.

cd release/pressure-triplets-326k-v1
sha256sum -c SHA256SUMS
mkdir -p pressure_triplets
for shard in $(seq -f "%03g" 0 31); do
  cat pressure-triplets-326k-v1-shard${shard}.tar.zst.part-* > pressure-triplets-326k-v1-shard${shard}.tar.zst
done
sha256sum -c SHA256SUMS.full
for archive in pressure-triplets-326k-v1-shard*.tar.zst; do
  tar --use-compress-program=unzstd -xf "$archive" -C pressure_triplets
done
gzip -dk pressure-triplets-326k-v1-manifest.tsv.gz

The triplet archives extract to paired NumPy arrays named:

<stem>_sad.npy
<stem>_elf.npy
<stem>_sym.npy

Current ELFNet training consumes _sad.npy and _elf.npy; _sym.npy is kept for provenance and compatibility with older SAD2ELF experiments.

The 75k DFT reference archive extracts selected reference ELFCAR files under paths matching the source Yuanhui structure tree. Use its manifest to map each archive path back to the selected-set rank, prototype, substituted elements, and ELF maxima values.