abstracts_embeddings - Sharded DiskANN Indices

Pre-built DiskANN indices sharded for distributed vector search.

Dataset Info

• Source: colonelwatch/abstracts-embeddings
• Vectors: 10,000,000
• Dimensions: 1024 (float32)
• Queries: 10,000

DiskANN Parameters

• R (graph degree): 64
• L (build beam width): 100
• Distance: L2

Shard Configurations

• shard_10: 10 shards x 1,000,000 vectors
• shard_3: 3 shards x 3,333,333 vectors
• shard_5: 5 shards x 2,000,000 vectors
• shard_7: 7 shards x 1,428,571 vectors

File Structure

fbin/
  base.fbin          # Base vectors (10,000,000 x 1024 x float32)
  queries.fbin       # Query vectors (10K x 1024 x float32)
diskann/
  shard_N/           # N-shard configuration
    *_disk.index                    # DiskANN disk index
    *_disk.index_512_none.indices   # MinIO graph indices
    *_disk.index_base_none.vectors  # MinIO vector data
    *_base.shardX.fbin              # Shard base data

Usage

Download with huggingface_hub

from huggingface_hub import hf_hub_download

# Download a specific shard configuration
index = hf_hub_download(
    repo_id="makneeeee/abstracts_embeddings",
    filename="diskann/shard_3/abstracts_embeddings_64_100_256.shard0_disk.index",
    repo_type="dataset"
)

Download with git-lfs

git lfs install
git clone https://huggingface.co/datasets/makneeeee/abstracts_embeddings

License

Same as source dataset.