Merge branch 'main' into please_dont_modify_this_branch_unless_you_ar…

…e_just_merging_with_main_

test/test_io.py

@@ -260,17 +260,14 @@ def test_write_video_with_audio(self, device, tmpdir):
         f_name = os.path.join(VIDEO_DIR, "R6llTwEh07w.mp4")
         video_tensor, audio_tensor, info = io.read_video(f_name, pts_unit="sec")
 
-        video_tensor = video_tensor.to(device)
-        audio_tensor = audio_tensor.to(device)
-
         out_f_name = os.path.join(tmpdir, "testing.mp4")
         io.video.write_video(
             out_f_name,
-            video_tensor,
+            video_tensor.to(device),
             round(info["video_fps"]),
             video_codec="libx264rgb",
             options={"crf": "0"},
-            audio_array=audio_tensor,
+            audio_array=audio_tensor.to(device),
             audio_fps=info["audio_fps"],
             audio_codec="aac",
         )

torchvision/datasets/kinetics.py

@@ -16,7 +16,7 @@
 from .vision import VisionDataset
 
 
-def _dl_wrap(tarpath: str, videopath: str, line: str) -> None:
+def _dl_wrap(tarpath: Union[str, Path], videopath: Union[str, Path], line: str) -> None:
     download_and_extract_archive(line, tarpath, videopath)
 
 

torchvision/datasets/utils.py

@@ -26,7 +26,7 @@
 
 def _urlretrieve(url: str, filename: Union[str, pathlib.Path], chunk_size: int = 1024 * 32) -> None:
     with urllib.request.urlopen(urllib.request.Request(url, headers={"User-Agent": USER_AGENT})) as response:
-        with open(filename, "wb") as fh, tqdm(total=response.length) as pbar:
+        with open(filename, "wb") as fh, tqdm(total=response.length, unit="B", unit_scale=True) as pbar:
             while chunk := response.read(chunk_size):
                 fh.write(chunk)
                 pbar.update(len(chunk))

torchvision/io/video.py

@@ -115,7 +115,7 @@ def write_video(
             audio_sample_fmt = container.streams.audio[0].format.name
 
             format_dtype = np.dtype(audio_format_dtypes[audio_sample_fmt])
-            audio_array = torch.as_tensor(audio_array).numpy().astype(format_dtype)
+            audio_array = torch.as_tensor(audio_array).numpy(force=True).astype(format_dtype)
 
             frame = av.AudioFrame.from_ndarray(audio_array, format=audio_sample_fmt, layout=audio_layout)
 

torchvision/transforms/_presets.py

@@ -2,7 +2,7 @@
 This file is part of the private API. Please do not use directly these classes as they will be modified on
 future versions without warning. The classes should be accessed only via the transforms argument of Weights.
 """
-from typing import Optional, Tuple
+from typing import Optional, Tuple, Union
 
 import torch
 from torch import nn, Tensor
@@ -87,7 +87,7 @@ def __init__(
         self,
         *,
         crop_size: Tuple[int, int],
-        resize_size: Tuple[int, int],
+        resize_size: Union[Tuple[int], Tuple[int, int]],
         mean: Tuple[float, ...] = (0.43216, 0.394666, 0.37645),
         std: Tuple[float, ...] = (0.22803, 0.22145, 0.216989),
         interpolation: InterpolationMode = InterpolationMode.BILINEAR,