Datasets:

lance-format
/

wiki-cohere-35m

Exception:    SplitsNotFoundError
Message:      The split names could not be parsed from the dataset config.
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
                  for split_generator in builder._split_generators(
                                         ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/src/libs/libcommon/src/libcommon/packaged_modules.py", line 18, in _refuse_lance
                  raise NotImplementedError(LANCE_DISABLED_MESSAGE)
              NotImplementedError: The Lance format is not supported.
              
              The above exception was the direct cause of the following exception:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 68, in compute_split_names_from_streaming_response
                  for split in get_dataset_split_names(
                               ~~~~~~~~~~~~~~~~~~~~~~~^
                      path=dataset,
                      ^^^^^^^^^^^^^
                      config_name=config,
                      ^^^^^^^^^^^^^^^^^^^
                      token=hf_token,
                      ^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
                  info = get_dataset_config_info(
                      path,
                  ...<6 lines>...
                      **config_kwargs,
                  )
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
                  raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
              datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.

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.

Wikipedia-Cohere 35M

35,000,000 base vectors and 5,000 query vectors, each with 768 float32 dimensions. This is the complete Wikipedia-Cohere 35M ANN benchmark, repackaged into Lance without changing or normalizing the vectors. It contains embeddings, not the original Wikipedia text.

Provenance and attribution

The Big ANN benchmark dataset definition identifies the base as Cohere embeddings of English Wikipedia titles and text, and the queries as Cohere embeddings of Simple English Wikipedia titles and text.

Downloaded source files:

The original embedding dataset was Cohere/wikipedia-22-12-en-embeddings. That upstream dataset card was unavailable when this conversion was prepared. This mirror does not assign a new license to the upstream vectors; consult the upstream owners for applicable terms. Attribution remains with Cohere, Wikipedia contributors, and the Big ANN benchmark maintainers.

The published scores preserve the source ground-truth values: negative inner products, where lower is better. They are not Lance's 1 - dot distances. Ground-truth IDs and ordering are preserved.

base.lance: original, unnormalized 768-dimensional float32 vectors in source order. Only the vector column is stored.
queries.lance: query_id (uint32) and vector (fixed-size list of 768 float32 values).
ground_truth.lance: query_id, neighbor_ids (100 uint64 values), and scores (100 values; semantics below).
SHA256SUMS: checksums of the data files and manifests.

All Lance tables are self-contained and have no vector or scalar indices. The base table contains only its initial index-free manifest, with no indexed versions or index files. No source .bin, .npy, .npz, pickle shards, or benchmark indices are included.

Ground-truth neighbor IDs are zero-based row positions in the original base table, suitable for base.take(ids). They are not passage IDs or a promise about Lance physical row IDs after rewriting the table.

The base data files use Lance file format 2.2; the small auxiliary tables use 2.0. Preparation was verified with pylance 13.0.0-beta.7. Use a current reader with file format 2.2 support.

Usage

import lance
from huggingface_hub import snapshot_download

root = snapshot_download("lance-format/wiki-cohere-35m", repo_type="dataset")
base = lance.dataset(f"{root}/base.lance")
queries = lance.dataset(f"{root}/queries.lance")
ground_truth = lance.dataset(f"{root}/ground_truth.lance")
assert base.count_rows() == 35000000
assert base.list_indices() == []

query = queries.take([0]).to_pylist()[0]["vector"]
expected = ground_truth.take([0]).to_pylist()[0]["neighbor_ids"]

The benchmark metric is maximum inner product (dot in Lance). Do not normalize vectors if reproducing this benchmark. Ties at the kth boundary may admit multiple correct neighbor sets.

Downloads last month: 185

Total file size:

108 GB

Wikipedia-Cohere 35M

Provenance and attribution

Contents

Usage