/data/vdb/cohere_large_10m/neighbors.parquet rows 1000 schema id: int64
neighbors_id: large_list<item: int64>
  child 0, item: int64
/data/vdb/cohere_large_10m/test.parquet rows 1000 schema id: int64
emb: large_list<item: float>
  child 0, item: float
/data/vdb/cohere_medium_1m/neighbors.parquet rows 1000 schema id: int64
neighbors_id: large_list<item: int64>
  child 0, item: int64
/data/vdb/cohere_medium_1m/test.parquet rows 1000 schema id: int64
emb: large_list<item: float>
  child 0, item: float
/data/vdb/cohere_medium_1m/train.parquet rows 1000000 schema id: int64
emb: large_list<item: float>
  child 0, item: float
/data/vdb/openai_large_5m/neighbors.parquet rows 1000 schema id: int64
neighbors_id: list<item: int64>
  child 0, item: int64
-- schema metadata --
pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 491
/data/vdb/openai_large_5m/test.parquet rows 1000 schema id: int64
emb: large_list<item: double>
  child 0, item: double
cohere_medium_1m: test (1000, 768) train (1000000, 768)
norms test min/max 12.342849 16.224667 train min/max 12.258187 17.474176
corpus exact-dup rows (after normalise): 81  test queries exactly equal to a corpus row: 1
test queries with max cos >= 1-1e-6: 1  >= 0.9999: 1
