Replace 'decord' with 'av' in VideoClassificationPipeline

src/transformers/__init__.py

@@ -1083,6 +1083,7 @@
         "add_end_docstrings",
         "add_start_docstrings",
         "is_apex_available",
+        "is_av_available",
         "is_bitsandbytes_available",
         "is_datasets_available",
         "is_decord_available",
@@ -5952,6 +5953,7 @@
         add_end_docstrings,
         add_start_docstrings,
         is_apex_available,
+        is_av_available,
         is_bitsandbytes_available,
         is_datasets_available,
         is_decord_available,

src/transformers/pipelines/video_classification.py

@@ -3,13 +3,19 @@
 
 import requests
 
-from ..utils import add_end_docstrings, is_decord_available, is_torch_available, logging, requires_backends
+from ..utils import (
+    add_end_docstrings,
+    is_av_available,
+    is_torch_available,
+    logging,
+    requires_backends,
+)
 from .base import Pipeline, build_pipeline_init_args
 
 
-if is_decord_available():
+if is_av_available():
+    import av
     import numpy as np
-    from decord import VideoReader
 
 
 if is_torch_available():
@@ -33,7 +39,7 @@ class VideoClassificationPipeline(Pipeline):
 
     def __init__(self, *args, **kwargs):
         super().__init__(*args, **kwargs)
-        requires_backends(self, "decord")
+        requires_backends(self, "av")
         self.check_model_type(MODEL_FOR_VIDEO_CLASSIFICATION_MAPPING_NAMES)
 
     def _sanitize_parameters(self, top_k=None, num_frames=None, frame_sampling_rate=None):
@@ -90,14 +96,13 @@ def preprocess(self, video, num_frames=None, frame_sampling_rate=1):
         if video.startswith("http://") or video.startswith("https://"):
             video = BytesIO(requests.get(video).content)
 
-        videoreader = VideoReader(video)
-        videoreader.seek(0)
+        container = av.open(video)
 
         start_idx = 0
         end_idx = num_frames * frame_sampling_rate - 1
         indices = np.linspace(start_idx, end_idx, num=num_frames, dtype=np.int64)
 
-        video = videoreader.get_batch(indices).asnumpy()
+        video = read_video_pyav(container, indices)
         video = list(video)
 
         model_inputs = self.image_processor(video, return_tensors=self.framework)
@@ -120,3 +125,16 @@ def postprocess(self, model_outputs, top_k=5):
         scores = scores.tolist()
         ids = ids.tolist()
         return [{"score": score, "label": self.model.config.id2label[_id]} for score, _id in zip(scores, ids)]
+
+
+def read_video_pyav(container, indices):
+    frames = []
+    container.seek(0)
+    start_index = indices[0]
+    end_index = indices[-1]
+    for i, frame in enumerate(container.decode(video=0)):
+        if i > end_index:
+            break
+        if i >= start_index and i in indices:
+            frames.append(frame)
+    return np.stack([x.to_ndarray(format="rgb24") for x in frames])

src/transformers/testing_utils.py

@@ -57,6 +57,7 @@
     is_aqlm_available,
     is_auto_awq_available,
     is_auto_gptq_available,
+    is_av_available,
     is_bitsandbytes_available,
     is_bs4_available,
     is_cv2_available,
@@ -1010,6 +1011,13 @@ def require_aqlm(test_case):
     return unittest.skipUnless(is_aqlm_available(), "test requires aqlm")(test_case)
 
 
+def require_av(test_case):
+    """
+    Decorator marking a test that requires av
+    """
+    return unittest.skipUnless(is_av_available(), "test requires av")(test_case)
+
+
 def require_bitsandbytes(test_case):
     """
     Decorator marking a test that requires the bitsandbytes library. Will be skipped when the library or its hard dependency torch is not installed.

src/transformers/utils/__init__.py

@@ -109,6 +109,7 @@
     is_aqlm_available,
     is_auto_awq_available,
     is_auto_gptq_available,
+    is_av_available,
     is_bitsandbytes_available,
     is_bs4_available,
     is_coloredlogs_available,

src/transformers/utils/import_utils.py

@@ -94,6 +94,7 @@ def _is_package_available(pkg_name: str, return_version: bool = False) -> Union[
 _accelerate_available, _accelerate_version = _is_package_available("accelerate", return_version=True)
 _apex_available = _is_package_available("apex")
 _aqlm_available = _is_package_available("aqlm")
+_av_available = importlib.util.find_spec("av") is not None
 _bitsandbytes_available = _is_package_available("bitsandbytes")
 _galore_torch_available = _is_package_available("galore_torch")
 # `importlib.metadata.version` doesn't work with `bs4` but `beautifulsoup4`. For `importlib.util.find_spec`, reversed.
@@ -656,6 +657,10 @@ def is_aqlm_available():
     return _aqlm_available
 
 
+def is_av_available():
+    return _av_available
+
+
 def is_ninja_available():
     r"""
     Code comes from *torch.utils.cpp_extension.is_ninja_available()*. Returns `True` if the
@@ -1012,6 +1017,16 @@ def is_mlx_available():
     return _mlx_available
 
 
+# docstyle-ignore
+AV_IMPORT_ERROR = """
+{0} requires the PyAv library but it was not found in your environment. You can install it with:
+```
+pip install av
+```
+Please note that you may need to restart your runtime after installation.
+"""
+
+
 # docstyle-ignore
 CV2_IMPORT_ERROR = """
 {0} requires the OpenCV library but it was not found in your environment. You can install it with:
@@ -1336,6 +1351,7 @@ def is_mlx_available():
 
 BACKENDS_MAPPING = OrderedDict(
     [
+        ("av", (is_av_available, AV_IMPORT_ERROR)),
         ("bs4", (is_bs4_available, BS4_IMPORT_ERROR)),
         ("cv2", (is_cv2_available, CV2_IMPORT_ERROR)),
         ("datasets", (is_datasets_available, DATASETS_IMPORT_ERROR)),

tests/pipelines/test_pipelines_video_classification.py

@@ -21,7 +21,7 @@
 from transformers.testing_utils import (
     is_pipeline_test,
     nested_simplify,
-    require_decord,
+    require_av,
     require_tf,
     require_torch,
     require_torch_or_tf,
@@ -34,7 +34,7 @@
 @is_pipeline_test
 @require_torch_or_tf
 @require_vision
-@require_decord
+@require_av
 class VideoClassificationPipelineTests(unittest.TestCase):
     model_mapping = MODEL_FOR_VIDEO_CLASSIFICATION_MAPPING