Merge remote-tracking branch 'origin/main' into user/aliberts/2025_02_25_refactor_robots

lerobot/common/datasets/lerobot_dataset.py

@@ -69,6 +69,7 @@ from lerobot.common.datasets.video_utils import (
     VideoFrame,
     decode_video_frames,
     encode_video_frames,
+    get_safe_default_codec,
     get_video_info,
 )
 from lerobot.common.robots.utils import Robot
@@ -462,7 +463,7 @@ class LeRobotDataset(torch.utils.data.Dataset):
             download_videos (bool, optional): Flag to download the videos. Note that when set to True but the
                 video files are already present on local disk, they won't be downloaded again. Defaults to
                 True.
-            video_backend (str | None, optional): Video backend to use for decoding videos. Defaults to torchcodec.
+            video_backend (str | None, optional): Video backend to use for decoding videos. Defaults to torchcodec when available int the platform; otherwise, defaults to 'pyav'.
                 You can also use the 'pyav' decoder used by Torchvision, which used to be the default option, or 'video_reader' which is another decoder of Torchvision.
         """
         super().__init__()
@@ -473,7 +474,7 @@ class LeRobotDataset(torch.utils.data.Dataset):
         self.episodes = episodes
         self.tolerance_s = tolerance_s
         self.revision = revision if revision else CODEBASE_VERSION
-        self.video_backend = video_backend if video_backend else "torchcodec"
+        self.video_backend = video_backend if video_backend else get_safe_default_codec()
         self.delta_indices = None
 
         # Unused attributes
@@ -1027,7 +1028,7 @@ class LeRobotDataset(torch.utils.data.Dataset):
         obj.delta_timestamps = None
         obj.delta_indices = None
         obj.episode_data_index = None
-        obj.video_backend = video_backend if video_backend is not None else "torchcodec"
+        obj.video_backend = video_backend if video_backend is not None else get_safe_default_codec()
         return obj
 
 

lerobot/common/datasets/video_utils.py

@@ -13,6 +13,7 @@
 # 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 importlib
 import json
 import logging
 import subprocess
@@ -27,14 +28,23 @@ import torch
 import torchvision
 from datasets.features.features import register_feature
 from PIL import Image
-from torchcodec.decoders import VideoDecoder
+
+
+def get_safe_default_codec():
+    if importlib.util.find_spec("torchcodec"):
+        return "torchcodec"
+    else:
+        logging.warning(
+            "'torchcodec' is not available in your platform, falling back to 'pyav' as a default decoder"
+        )
+        return "pyav"
 
 
 def decode_video_frames(
     video_path: Path | str,
     timestamps: list[float],
     tolerance_s: float,
-    backend: str = "torchcodec",
+    backend: str | None = None,
 ) -> torch.Tensor:
     """
     Decodes video frames using the specified backend.
@@ -43,13 +53,15 @@ def decode_video_frames(
         video_path (Path): Path to the video file.
         timestamps (list[float]): List of timestamps to extract frames.
         tolerance_s (float): Allowed deviation in seconds for frame retrieval.
-        backend (str, optional): Backend to use for decoding. Defaults to "torchcodec".
+        backend (str, optional): Backend to use for decoding. Defaults to "torchcodec" when available in the platform; otherwise, defaults to "pyav"..
 
     Returns:
         torch.Tensor: Decoded frames.
 
     Currently supports torchcodec on cpu and pyav.
     """
+    if backend is None:
+        backend = get_safe_default_codec()
     if backend == "torchcodec":
         return decode_video_frames_torchcodec(video_path, timestamps, tolerance_s)
     elif backend in ["pyav", "video_reader"]:
@@ -173,6 +185,12 @@ def decode_video_frames_torchcodec(
     and all subsequent frames until reaching the requested frame. The number of key frames in a video
     can be adjusted during encoding to take into account decoding time and video size in bytes.
     """
+
+    if importlib.util.find_spec("torchcodec"):
+        from torchcodec.decoders import VideoDecoder
+    else:
+        raise ImportError("torchcodec is required but not available.")
+
     # initialize video decoder
     decoder = VideoDecoder(video_path, device=device, seek_mode="approximate")
     loaded_frames = []