The dataset viewer is not available for this split.
Error code: RowsPostProcessingError
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
Processed BONES-SEED SMPL–G1 pairs
Existing processed NPZ clips packaged into uncompressed tar shards. No motion conversion is performed during packaging.
Download and extract from the project root:
hf download NONHUMAN-RESEARCH/seed-processed-full --repo-type dataset --include "shards/*" --include "shards.json" --include "manifest.jsonl" --include "preprocess.json" --include "summary.json" --local-dir data/seed-processed-full-download
mkdir -p data/seed-processed-full
for shard in data/seed-processed-full-download/shards/*/*/*.tar; do
tar -xf "$shard" -C data/seed-processed-full
done
cp data/seed-processed-full-download/{manifest.jsonl,preprocess.json,summary.json} data/seed-processed-full/
Extraction restores train/ and validation/ for the existing loaders. shards.json records clip counts and SHA-256 checksums. The original preprocessing metadata and manifest are preserved; source paths in them describe the machine that processed the data.
Shards are grouped in directories of at most 100 archives. Training accepts the downloaded package root with bin/train.sh --data PATH. Evaluation accepts --data PATH/validation; archives are verified and extracted once into PATH/.unpacked (requires additional disk space).
- Downloads last month
- 82