fix(tests)

- Updated `lerobot_dataset.py:add_frame` to take task as key in frame
- Updated `lerobot_dataset.py` to remove robot argument from `create` function of lerobotdataset and lerobotdatasetmetadata and directly take the features
- Update `test_datasets.py` to features from Mock robot
- Update all the usage of `add_frame` in the library
- Update `dataset_factories.py`; had issues with new argument order
- Raise ValueError when no task is provided (in `datasets/utils.py` validate func)

src/lerobot/record.py

@@ -218,8 +218,8 @@ def record_loop(
 
         if dataset is not None:
             action_frame = build_dataset_frame(dataset.features, sent_action, prefix="action")
-            frame = {**observation_frame, **action_frame}
-            dataset.add_frame(frame, task=single_task)
+            frame = {**observation_frame, **action_frame, "task": single_task}
+            dataset.add_frame(frame)
 
         if display_data:
             for obs, val in observation.items():

src/lerobot/scripts/rl/crop_dataset_roi.py

@@ -227,7 +227,8 @@ def convert_lerobot_dataset_to_cropper_lerobot_dataset(
 
             new_frame[key] = value
 
-        new_dataset.add_frame(new_frame, task=task)
+        new_frame["task"] = task
+        new_dataset.add_frame(new_frame)
 
         if frame["episode_index"].item() != prev_episode_index:
             # Save the episode

src/lerobot/scripts/rl/gym_manipulator.py

@@ -2132,7 +2132,8 @@ def record_dataset(env, policy, cfg):
             frame["complementary_info.discrete_penalty"] = torch.tensor(
                 [info.get("discrete_penalty", 0.0)], dtype=torch.float32
             )
-            dataset.add_frame(frame, task=cfg.task)
+            frame["task"] = cfg.task
+            dataset.add_frame(frame)
 
             # Maintain consistent timing
             if cfg.fps:

src/lerobot/utils/buffer.py

@@ -596,7 +596,8 @@ class ReplayBuffer:
                         frame_dict[f"complementary_info.{key}"] = val
 
             # Add to the dataset's buffer
-            lerobot_dataset.add_frame(frame_dict, task=task_name)
+            frame_dict["task"] = task_name
+            lerobot_dataset.add_frame(frame_dict)
 
             # Move to next frame
             frame_idx_in_episode += 1