Buckets:

glennmatlin's picture
download
raw
1.81 kB
from __future__ import annotations
import json
import sys
from pathlib import Path
import numpy as np
import pytest
REPO_ROOT = Path(__file__).resolve().parents[1]
QUERY_DATA_DIR = REPO_ROOT / "scripts" / "query_data"
if str(QUERY_DATA_DIR) not in sys.path:
sys.path.insert(0, str(QUERY_DATA_DIR))
import merge_scores # noqa: E402
def _write_score_dir(path: Path, values: list[list[float]]) -> None:
num_scores = len(values)
num_items = len(values[0])
dtype_json = {
"names": [
item for i in range(num_scores) for item in (f"score_{i}", f"written_{i}")
],
"formats": [item for _ in range(num_scores) for item in ("float32", "bool")],
"offsets": [item for i in range(num_scores) for item in (i * 8, i * 8 + 4)],
"itemsize": num_scores * 8,
}
dtype = np.dtype(dtype_json)
mmap = np.memmap(path / "scores.bin", dtype=dtype, mode="w+", shape=(num_items,))
for index, column in enumerate(values):
mmap[f"score_{index}"] = np.array(column, dtype=np.float32)
mmap[f"written_{index}"] = True
mmap.flush()
info = {
"num_items": num_items,
"num_scores": num_scores,
"dtype": dtype_json,
}
(path / "info.json").write_text(json.dumps(info), encoding="utf-8")
def test_merge_scores_load_fp32_scores(tmp_path: Path) -> None:
_write_score_dir(tmp_path, [[1.5, 3.0, -2.0]])
merge_loaded, num_items = merge_scores.load_scores(tmp_path)
assert num_items == 3
assert merge_loaded.tolist() == pytest.approx([1.5, 3.0, -2.0])
def test_merge_scores_reject_multi_score_outputs(tmp_path: Path) -> None:
_write_score_dir(tmp_path, [[1.0, 2.0], [3.0, 4.0]])
with pytest.raises(ValueError, match="expected 1"):
merge_scores.load_scores(tmp_path)

Xet Storage Details

Size:
1.81 kB
·
Xet hash:
91f4a85f547a71ead497955e805c566d4815d4f3fa583595b05c1d7ca9d1154f

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.