Refactored clip_sampler (#1562)

torchvision/datasets/samplers/clip_sampler.py

@@ -2,7 +2,7 @@ import math
 import torch
 from torch.utils.data import Sampler
 import torch.distributed as dist
-import torchvision.datasets.video_utils
+from torchvision.datasets.video_utils import VideoClips
 
 
 class DistributedSampler(Sampler):
@@ -96,7 +96,7 @@ class DistributedSampler(Sampler):
         self.epoch = epoch
 
 
-class UniformClipSampler(torch.utils.data.Sampler):
+class UniformClipSampler(Sampler):
     """
     Sample `num_video_clips_per_video` clips for each video, equally spaced.
     When number of unique clips in the video is fewer than num_video_clips_per_video,
@@ -107,7 +107,7 @@ class UniformClipSampler(torch.utils.data.Sampler):
         num_clips_per_video (int): number of clips to be sampled per video
     """
     def __init__(self, video_clips, num_clips_per_video):
-        if not isinstance(video_clips, torchvision.datasets.video_utils.VideoClips):
+        if not isinstance(video_clips, VideoClips):
             raise TypeError("Expected video_clips to be an instance of VideoClips, "
                             "got {}".format(type(video_clips)))
         self.video_clips = video_clips
@@ -139,7 +139,7 @@ class UniformClipSampler(torch.utils.data.Sampler):
         )
 
 
-class RandomClipSampler(torch.utils.data.Sampler):
+class RandomClipSampler(Sampler):
     """
     Samples at most `max_video_clips_per_video` clips for each video randomly
 
@@ -148,7 +148,7 @@ class RandomClipSampler(torch.utils.data.Sampler):
         max_clips_per_video (int): maximum number of clips to be sampled per video
     """
     def __init__(self, video_clips, max_clips_per_video):
-        if not isinstance(video_clips, torchvision.datasets.video_utils.VideoClips):
+        if not isinstance(video_clips, VideoClips):
             raise TypeError("Expected video_clips to be an instance of VideoClips, "
                             "got {}".format(type(video_clips)))
         self.video_clips = video_clips