From e0391163d550c4a8c6804454f2c5d687fe5d6928 Mon Sep 17 00:00:00 2001 From: ahadjawaid <94938815+ahadjawaid@users.noreply.github.com> Date: Tue, 2 Dec 2025 08:40:11 -0800 Subject: [PATCH] Skip hidden .parquet files when loading dataset to avoid PyArrow errors Signed-off-by: ahadjawaid <94938815+ahadjawaid@users.noreply.github.com> --- src/lerobot/datasets/utils.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/lerobot/datasets/utils.py b/src/lerobot/datasets/utils.py index 234736a756..6f0673d558 100644 --- a/src/lerobot/datasets/utils.py +++ b/src/lerobot/datasets/utils.py @@ -124,7 +124,7 @@ def load_nested_dataset( # When no filtering needed, Dataset uses memory-mapped loading for efficiency # PyArrow loads the entire dataset into memory if episodes is None: - return Dataset.from_parquet([str(path) for path in paths], features=features) + return Dataset.from_parquet([str(path) for path in paths if not path.name.startswith(".")], features=features) arrow_dataset = pa_ds.dataset(paths, format="parquet") filter_expr = pa_ds.field("episode_index").isin(episodes)