Preserve private Hub auth in streaming readers

This commit is contained in:
Pepijn
2026-07-27 09:41:57 +02:00
parent b5a13e43ce
commit c0e9b0bbff
8 changed files with 74 additions and 10 deletions
@@ -11,6 +11,7 @@
from __future__ import annotations
from pathlib import Path
from unittest.mock import Mock
import fsspec
import pytest
@@ -108,3 +109,16 @@ def test_reader_supports_fsspec_remote_root() -> None:
table = reader.read_episode("data/chunk-000/file-000.parquet", episode_index=0, expected_rows=3)
assert table.column("value").to_pylist() == [0, 1, 2]
def test_reader_forwards_explicit_token_to_hf_filesystem(monkeypatch) -> None:
url_to_fs = Mock(return_value=(fsspec.filesystem("memory"), "datasets/private@revision"))
monkeypatch.setattr(fsspec.core, "url_to_fs", url_to_fs)
EpisodeParquetReader(
"hf://datasets/private@revision",
columns=("episode_index",),
token="hf_test_token",
)
url_to_fs.assert_called_once_with("hf://datasets/private@revision", token="hf_test_token")
+3
View File
@@ -46,7 +46,9 @@ def test_streaming_dataset_forwards_token_to_metadata_without_retaining_it(
rescale_depth_stats=Mock(),
)
metadata_cls = Mock(return_value=metadata)
ensure_sidecar = Mock(return_value=None)
monkeypatch.setattr(streaming_dataset_module, "LeRobotDatasetMetadata", metadata_cls)
monkeypatch.setattr(streaming_dataset_module, "ensure_dataset_mp4_sidecar", ensure_sidecar)
dataset = StreamingLeRobotDataset(DUMMY_REPO_ID, root=requested_root, token=token)
@@ -57,6 +59,7 @@ def test_streaming_dataset_forwards_token_to_metadata_without_retaining_it(
force_cache_sync=False,
token=token,
)
assert ensure_sidecar.call_args.kwargs["token"] is (None if from_local else token)
assert not hasattr(dataset, "_token")