fix a bug related to audio_end_pts (#1431)

Author: stephenyan1231

torchvision/datasets/video_utils.py

@@ -308,7 +308,7 @@ def get_clip(self, idx):
                     math.floor,
                 )
                 audio_end_pts = pts_convert(
-                    video_start_pts,
+                    video_end_pts,
                     info["video_timebase"],
                     info["audio_timebase"],
                     math.ceil,