fix(datasets): allow zero-width features in get_hf_features_from_features

Setting a 1-D feature with shape=(0,) builds datasets.Sequence(length=0, ...),
which pyarrow rejects with ArrowInvalid: list_size needs to be a strict
positive integer when datasets.Dataset.from_dict(...) is called inside
save_episode. Use length=-1 (variable-length) for zero-width 1-D shapes.

Fixes the second half of #3654 (the first half is #3664, in compute_episode_stats).

src/lerobot/datasets/feature_utils.py

@@ -67,9 +67,9 @@ def get_hf_features_from_features(features: dict) -> datasets.Features:
         elif ft["shape"] == (1,):
             hf_features[key] = datasets.Value(dtype=ft["dtype"])
         elif len(ft["shape"]) == 1:
-            hf_features[key] = datasets.Sequence(
-                length=ft["shape"][0], feature=datasets.Value(dtype=ft["dtype"])
-            )
+            # pyarrow rejects fixed_size_list[0], so use a variable length list instead
+            length = ft["shape"][0] if ft["shape"][0] > 0 else -1
+            hf_features[key] = datasets.Sequence(length=length, feature=datasets.Value(dtype=ft["dtype"]))
         elif len(ft["shape"]) == 2:
             hf_features[key] = datasets.Array2D(shape=ft["shape"], dtype=ft["dtype"])
         elif len(ft["shape"]) == 3:

tests/datasets/test_feature_utils.py

@@ -0,0 +1,29 @@
+#!/usr/bin/env python
+
+# Copyright 2026 The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import pytest
+
+pytest.importorskip("datasets", reason="datasets is required (install lerobot[dataset])")
+
+import datasets
+
+from lerobot.datasets.feature_utils import get_hf_features_from_features
+
+
+def test_get_hf_features_zero_width_feature_does_not_raise_on_from_dict():
+    features = {"empty": {"dtype": "float32", "shape": (0,), "names": ["empty"]}}
+    hf_features = get_hf_features_from_features(features)
+    datasets.Dataset.from_dict({"empty": [[], []]}, features=hf_features)