extend DistributedSampler to support group_size (#1512)

* extend DistributedSampler to support group_size

* Fix lint

Author: stephenyan1231

test/test_datasets_samplers.py

@@ -5,7 +5,11 @@
 import unittest
 
 from torchvision import io
-from torchvision.datasets.samplers import RandomClipSampler, UniformClipSampler
+from torchvision.datasets.samplers import (
+    DistributedSampler,
+    RandomClipSampler,
+    UniformClipSampler,
+)
 from torchvision.datasets.video_utils import VideoClips, unfold
 from torchvision import get_video_backend
 
@@ -83,6 +87,31 @@ def test_uniform_clip_sampler_insufficient_clips(self):
             indices = torch.tensor(list(iter(sampler)))
             self.assertTrue(indices.equal(torch.tensor([0, 0, 1, 2, 4, 6, 7, 9, 11])))
 
+    def test_distributed_sampler_and_uniform_clip_sampler(self):
+        with get_list_of_videos(num_videos=3, sizes=[25, 25, 25]) as video_list:
+            video_clips = VideoClips(video_list, 5, 5)
+            clip_sampler = UniformClipSampler(video_clips, 3)
+
+            distributed_sampler_rank0 = DistributedSampler(
+                clip_sampler,
+                num_replicas=2,
+                rank=0,
+                group_size=3,
+            )
+            indices = torch.tensor(list(iter(distributed_sampler_rank0)))
+            self.assertEqual(len(distributed_sampler_rank0), 6)
+            self.assertTrue(indices.equal(torch.tensor([0, 2, 4, 10, 12, 14])))
+
+            distributed_sampler_rank1 = DistributedSampler(
+                clip_sampler,
+                num_replicas=2,
+                rank=1,
+                group_size=3,
+            )
+            indices = torch.tensor(list(iter(distributed_sampler_rank1)))
+            self.assertEqual(len(distributed_sampler_rank1), 6)
+            self.assertTrue(indices.equal(torch.tensor([5, 7, 9, 0, 2, 4])))
+
 
 if __name__ == '__main__':
     unittest.main()

torchvision/datasets/samplers/clip_sampler.py

@@ -9,9 +9,32 @@ class DistributedSampler(Sampler):
     """
     Extension of DistributedSampler, as discussed in
     https://github.com/pytorch/pytorch/issues/23430
+
+    Example:
+        dataset: [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13]
+        num_replicas: 4
+        shuffle: False
+
+    when group_size = 1
+            RANK    |  shard_dataset
+            =========================
+            rank_0  |  [0, 4, 8, 12]
+            rank_1  |  [1, 5, 9, 13]
+            rank_2  |  [2, 6, 10, 0]
+            rank_3  |  [3, 7, 11, 1]
+
+    when group_size = 2
+
+            RANK    |  shard_dataset
+            =========================
+            rank_0  |  [0, 1, 8, 9]
+            rank_1  |  [2, 3, 10, 11]
+            rank_2  |  [4, 5, 12, 13]
+            rank_3  |  [6, 7, 0, 1]
+
     """
 
-    def __init__(self, dataset, num_replicas=None, rank=None, shuffle=False):
+    def __init__(self, dataset, num_replicas=None, rank=None, shuffle=False, group_size=1):
         if num_replicas is None:
             if not dist.is_available():
                 raise RuntimeError("Requires distributed package to be available")
@@ -20,11 +43,20 @@ def __init__(self, dataset, num_replicas=None, rank=None, shuffle=False):
             if not dist.is_available():
                 raise RuntimeError("Requires distributed package to be available")
             rank = dist.get_rank()
+        assert len(dataset) % group_size == 0, (
+            "dataset length must be a multiplier of group size"
+            "dataset length: %d, group size: %d" % (len(dataset), group_size)
+        )
         self.dataset = dataset
+        self.group_size = group_size
         self.num_replicas = num_replicas
         self.rank = rank
         self.epoch = 0
-        self.num_samples = int(math.ceil(len(self.dataset) * 1.0 / self.num_replicas))
+        dataset_group_length = len(dataset) // group_size
+        self.num_group_samples = int(
+            math.ceil(dataset_group_length * 1.0 / self.num_replicas)
+        )
+        self.num_samples = self.num_group_samples * group_size
         self.total_size = self.num_samples * self.num_replicas
         self.shuffle = shuffle
 
@@ -41,8 +73,14 @@ def __iter__(self):
         indices += indices[:(self.total_size - len(indices))]
         assert len(indices) == self.total_size
 
+        total_group_size = self.total_size // self.group_size
+        indices = torch.reshape(
+            torch.LongTensor(indices), (total_group_size, self.group_size)
+        )
+
         # subsample
-        indices = indices[self.rank:self.total_size:self.num_replicas]
+        indices = indices[self.rank:total_group_size:self.num_replicas, :]
+        indices = torch.reshape(indices, (-1,)).tolist()
         assert len(indices) == self.num_samples
 
         if isinstance(self.dataset, Sampler):