scripts/restore.sh

#!/usr/bin/env bash
# Download RNA-SBDD v2, verify every digest, and extract the dataset.
#
#   ./restore.sh /your/data/root
#
# Leaves you with:
#   $ROOT/rna-sbdd-v2-download/          the raw repository download
#   $ROOT/rna-sbdd-v2-<id>-model-ready/  the extracted dataset release
#   $ROOT/benchmark_artifacts/           evaluation artifacts
#   $ROOT/checkpoints/                   model weights, digest-verified
#
# It does NOT remap the repository's absolute paths; run scripts/remap_paths.py
# for that, after this finishes.

set -euo pipefail

ROOT="${1:?usage: restore.sh /your/data/root}"
REPO_ID="CedLJH/rna-sbdd-v2"
DOWNLOAD="$ROOT/rna-sbdd-v2-download"

command -v hf >/dev/null || {
    echo "hf CLI not found: pip install -U huggingface_hub" >&2; exit 1; }
command -v unzstd >/dev/null || {
    echo "unzstd not found: install zstd" >&2; exit 1; }

mkdir -p "$ROOT"

echo "==> downloading $REPO_ID"
hf download "$REPO_ID" --repo-type=dataset --local-dir "$DOWNLOAD"

echo "==> verifying the dataset archive"
( cd "$DOWNLOAD/dataset" && sha256sum -c SHA256SUMS )

echo "==> extracting the dataset release"
tar --use-compress-program=unzstd \
    -xf "$DOWNLOAD"/dataset/*.tar.zst -C "$ROOT"

echo "==> placing benchmark artifacts"
rm -rf "$ROOT/benchmark_artifacts"
cp -r "$DOWNLOAD/benchmark_artifacts" "$ROOT/benchmark_artifacts"

echo "==> placing checkpoints"
rm -rf "$ROOT/checkpoints"
cp -r "$DOWNLOAD/checkpoints" "$ROOT/checkpoints"

echo "==> verifying checkpoint digests"
python3 - "$ROOT/checkpoints" <<'PY'
import hashlib, json, sys
from pathlib import Path

root = Path(sys.argv[1])
manifest = json.loads((root / "MANIFEST.json").read_text())
entries = manifest["bound"] + manifest["merged_wave_terminal"]

bad = missing = 0
for entry in entries:
    path = root.parent / entry["repo_path"]
    if not path.is_file():
        print(f"  MISSING  {entry['repo_path']}")
        missing += 1
        continue
    digest = hashlib.sha256()
    with open(path, "rb") as handle:
        for block in iter(lambda: handle.read(1 << 20), b""):
            digest.update(block)
    if digest.hexdigest() != entry["sha256"]:
        print(f"  CORRUPT  {entry['repo_path']}")
        bad += 1

total = len(entries)
print(f"  {total - bad - missing}/{total} checkpoints verified")
if bad or missing:
    sys.exit(1)
PY

cat <<EOF

Done.  Dataset, artifacts, and weights are in place under $ROOT.

Next, bind the repository to these paths:

  python scripts/remap_paths.py --repo /path/to/RNASBDD \\
      --data-root        $ROOT \\
      --env-prefix       /your/conda/envs/rnasbdd \\
      --crossdocked-root /your/sbdd_data

That is a dry run until you add --apply.  Afterwards, run the test suite.
EOF