Bug: Incomplete Dataset for yandex-1B-200-angular/yandex_t2i_gt_100k

What is the issue?

I am trying to run the benchmark on my own adding in a new vector database LanceDB specifically to compare filter search performance. When I try running the benchmark using this command:

poetry run python3 run.py --engines "lancedb-*"

I get the following error:

FileNotFoundError: [Errno 2] No such file or directory: '/Users/Yudhiesh/Projects/vector-db-benchmark/datasets/yandex-1B-200-angular/yandex_t2i_gt_100k/vectors.npy'

I noticed that the dataset that is downloaded is incomplete:

datasets/yandex-1B-200-angular
└── yandex_t2i_gt_100k
    └── tests.jsonl

2 directories, 1 file

Based on the specification of the AnnCompoundReader since the files are .jsonl it is missing the actual vectors:

class AnnCompoundReader(JSONReader):
    """
    A reader created specifically to read the format used in
    https://github.com/qdrant/ann-filtering-benchmark-datasets, in which vectors
    and their metadata are stored in separate files.
    """

    VECTORS_FILE = "vectors.npy"
    QUERIES_FILE = "tests.jsonl"

Start with run.py and the AnnCompoundReader definition, then inspect how yandex-1B-200-angular/yandex_t2i_gt_100k is downloaded or assembled. Reproduce the command with the lancedb-* engine pattern and verify that the dataset includes the vectors.npy file required by the reader and that the benchmark proceeds past dataset loading.