Merge branch 'main' into feat/implement_evo1

tests/scripts/test_edit_dataset_parsing.py

@@ -27,6 +27,7 @@ from lerobot.scripts.lerobot_edit_dataset import (
     MergeConfig,
     ModifyTasksConfig,
     OperationConfig,
+    ReencodeVideosConfig,
     RemoveFeatureConfig,
     SplitConfig,
     _validate_config,
@@ -66,6 +67,20 @@ class TestOperationTypeParsing:
         with pytest.raises(ValueError, match="--new_repo_id is required for merge"):
             _validate_config(cfg)
 
+    @pytest.mark.parametrize("flag", ["concatenate_videos", "concatenate_data"])
+    def test_merge_concatenate_flag_defaults_true(self, flag):
+        cfg = parse_cfg(["--new_repo_id", "test/merged", "--operation.type", "merge"])
+        assert isinstance(cfg.operation, MergeConfig)
+        assert getattr(cfg.operation, flag) is True
+
+    @pytest.mark.parametrize("flag", ["concatenate_videos", "concatenate_data"])
+    def test_merge_concatenate_flag_can_be_disabled(self, flag):
+        cfg = parse_cfg(
+            ["--new_repo_id", "test/merged", "--operation.type", "merge", f"--operation.{flag}", "false"]
+        )
+        assert isinstance(cfg.operation, MergeConfig)
+        assert getattr(cfg.operation, flag) is False
+
     def test_non_merge_requires_repo_id(self):
         cfg = parse_cfg(["--operation.type", "delete_episodes"])
         with pytest.raises(ValueError, match="--repo_id is required for delete_episodes"):
@@ -89,3 +104,47 @@ class TestOperationTypeParsing:
         )
         resolved_name = OperationConfig.get_choice_name(type(cfg.operation))
         assert resolved_name == type_name
+
+
+class TestDepthEncoderParsing:
+    """Test that the depth encoder is exposed and parsed for video operations."""
+
+    def test_reencode_has_default_depth_encoder(self):
+        cfg = parse_cfg(["--repo_id", "test/repo", "--operation.type", "reencode_videos"])
+        assert isinstance(cfg.operation, ReencodeVideosConfig)
+        # A depth encoder is configured by default so depth videos are re-encoded too.
+        assert cfg.operation.depth_encoder is not None
+        assert hasattr(cfg.operation.depth_encoder, "depth_min")
+
+    def test_reencode_parses_depth_encoder_overrides(self):
+        cfg = parse_cfg(
+            [
+                "--repo_id",
+                "test/repo",
+                "--operation.type",
+                "reencode_videos",
+                "--operation.depth_encoder.extra_options",
+                '{"x265-params": "lossless=1"}',
+                "--operation.depth_encoder.depth_max",
+                "12.0",
+                "--operation.depth_encoder.use_log",
+                "false",
+            ]
+        )
+        assert cfg.operation.depth_encoder.extra_options == {"x265-params": "lossless=1"}
+        assert cfg.operation.depth_encoder.depth_max == 12.0
+        assert cfg.operation.depth_encoder.use_log is False
+
+    def test_convert_image_to_video_parses_depth_encoder_overrides(self):
+        cfg = parse_cfg(
+            [
+                "--repo_id",
+                "test/repo",
+                "--operation.type",
+                "convert_image_to_video",
+                "--operation.depth_encoder.depth_min",
+                "0.05",
+            ]
+        )
+        assert isinstance(cfg.operation, ConvertImageToVideoConfig)
+        assert cfg.operation.depth_encoder.depth_min == 0.05

tests/scripts/test_lerobot_annotate.py

@@ -0,0 +1,86 @@
+#!/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 json
+from types import SimpleNamespace
+
+import pytest
+
+# ``lerobot.scripts.lerobot_annotate`` (and the ``_push_to_hub`` path it
+# exercises) imports ``lerobot.datasets``, which only ships under the
+# ``dataset`` extra. Skip in tiers without it instead of erroring.
+pytest.importorskip("datasets", reason="datasets is required (install lerobot[dataset])")
+
+
+def test_push_to_hub_tags_uploaded_dataset_revision(tmp_path, monkeypatch):
+    from lerobot.scripts.lerobot_annotate import _push_to_hub
+
+    root = tmp_path / "dataset"
+    (root / "meta").mkdir(parents=True)
+    (root / "meta" / "info.json").write_text(json.dumps({"codebase_version": "v3.0"}))
+
+    calls = {}
+
+    class FakeHfApi:
+        def create_repo(self, **kwargs):
+            calls["create_repo"] = kwargs
+
+        def upload_folder(self, **kwargs):
+            calls["upload_folder"] = kwargs
+            return SimpleNamespace(oid="abc123")
+
+        def delete_tag(self, repo_id, **kwargs):
+            import requests
+            from huggingface_hub.errors import RevisionNotFoundError
+
+            calls["delete_tag"] = {"repo_id": repo_id, **kwargs}
+            # Simulate the common case: no stale tag to delete.
+            raise RevisionNotFoundError("no such tag", response=requests.Response())
+
+        def create_tag(self, **kwargs):
+            calls["create_tag"] = kwargs
+
+    monkeypatch.setattr("huggingface_hub.HfApi", FakeHfApi)
+
+    cfg = SimpleNamespace(
+        repo_id="source/dataset",
+        new_repo_id="annotated/dataset",
+        push_private=True,
+        push_commit_message=None,
+    )
+
+    _push_to_hub(root, cfg)
+
+    assert calls["create_repo"] == {
+        "repo_id": "annotated/dataset",
+        "repo_type": "dataset",
+        "private": True,
+        "exist_ok": True,
+    }
+    assert calls["upload_folder"]["repo_id"] == "annotated/dataset"
+    # A stale tag (e.g. from a previous annotation run) is deleted first so
+    # the new tag always points at the upload we just made.
+    assert calls["delete_tag"] == {
+        "repo_id": "annotated/dataset",
+        "tag": "v3.0",
+        "repo_type": "dataset",
+    }
+    assert calls["create_tag"] == {
+        "repo_id": "annotated/dataset",
+        "tag": "v3.0",
+        "repo_type": "dataset",
+        "revision": "abc123",
+    }

tests/scripts/test_train_remote_dispatch.py

@@ -0,0 +1,67 @@
+# Copyright 2025 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 sys
+
+import draccus
+import pytest
+
+# Importing lerobot_train eagerly pulls in lerobot.datasets, which needs the
+# `dataset` extra. The base CI tier runs without it, so skip the whole module there.
+pytest.importorskip("datasets", reason="datasets is required (install lerobot[dataset])")
+
+from lerobot.configs.train import TrainPipelineConfig  # noqa: E402
+from lerobot.policies.act.configuration_act import (
+    ACTConfig,  # noqa: E402, F401  (registers --policy.type act)
+)
+from lerobot.scripts.lerobot_train import _remote_target_in_argv, train  # noqa: E402
+
+
+def _set_argv(monkeypatch, *args):
+    monkeypatch.setattr(sys, "argv", ["lerobot-train", *args])
+
+
+def test_remote_target_detected_space_separated(monkeypatch):
+    _set_argv(monkeypatch, "--policy.type", "act", "--job.target", "a10g-small")
+    assert _remote_target_in_argv() is True
+
+
+def test_remote_target_detected_equals(monkeypatch):
+    _set_argv(monkeypatch, "--job.target=t4-small")
+    assert _remote_target_in_argv() is True
+
+
+def test_local_string_is_not_remote(monkeypatch):
+    _set_argv(monkeypatch, "--job.target", "local")
+    assert _remote_target_in_argv() is False
+
+
+def test_no_target_is_not_remote(monkeypatch):
+    _set_argv(monkeypatch, "--policy.type", "act")
+    assert _remote_target_in_argv() is False
+
+
+def test_train_dispatches_to_submit_when_remote(monkeypatch):
+    """A remote --job.target short-circuits train() to the HF Jobs submitter."""
+    import lerobot.scripts.lerobot_train as train_module
+
+    captured = []
+    monkeypatch.setattr(train_module, "submit_to_hf", lambda cfg: captured.append(cfg) or "submitted")
+    cfg = draccus.parse(
+        TrainPipelineConfig,
+        args=["--dataset.repo_id", "u/d", "--policy.type", "act", "--job.target", "a10g-small"],
+    )
+    # Returns the submitter's result and never enters the local training path.
+    assert train(cfg) == "submitted"
+    assert captured == [cfg]