[bugfix] Add support for CombinedLoader in validation with ddp

CHANGELOG.md

@@ -301,6 +301,9 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Fixed metric objects passed directly to `self.log` not being reset correctly ([#7055](https://github.com/PyTorchLightning/pytorch-lightning/pull/7055))
 
 
+- Fixed `CombinedLoader` in distributed settings for validation / testing ([#7102](https://github.com/PyTorchLightning/pytorch-lightning/pull/7102))
+
+
 ## [1.2.7] - 2021-04-06
 
 ### Fixed

pytorch_lightning/trainer/data_loading.py

@@ -108,12 +108,15 @@ def auto_add_worker_init_fn(self, dataloader: DataLoader) -> None:
             dataloader.worker_init_fn = partial(pl_worker_init_function, rank=self.global_rank)
 
     def auto_add_sampler(self, dataloader: DataLoader, shuffle: bool) -> DataLoader:
-
         # don't do anything if it's not a dataloader
         is_dataloader = isinstance(dataloader, DataLoader)
         # don't manipulate iterable datasets
         is_iterable_ds = has_iterable_dataset(dataloader)
 
+        if isinstance(dataloader, CombinedLoader):
+            dataloader.loaders = apply_to_collection(dataloader.loaders, DataLoader, self.auto_add_sampler, shuffle)
+            return dataloader
+
         if not is_dataloader or is_iterable_ds:
             return dataloader
 

pytorch_lightning/trainer/supporters.py

@@ -19,6 +19,8 @@
 import torch
 from torch import Tensor
 from torch.utils.data import Dataset
+from torch.utils.data.dataloader import DataLoader
+from torch.utils.data.dataset import IterableDataset
 
 from pytorch_lightning.utilities.apply_func import apply_to_collection
 from pytorch_lightning.utilities.cloud_io import get_filesystem
@@ -352,7 +354,7 @@ def __init__(self, loaders: Any, mode: str = 'min_size'):
     @property
     def sampler(self) -> Union[Iterable, Sequence, Mapping]:
         """Return a collections of samplers extracting from loaders."""
-        return apply_to_collection(self.loaders, Iterable, getattr, 'sampler', None, wrong_dtype=(Sequence, Mapping))
+        return apply_to_collection(self.loaders, (DataLoader, IterableDataset), getattr, 'sampler', None)
 
     def _wrap_loaders_max_size_cycle(self) -> Any:
         """

tests/trainer/test_supporters.py

@@ -11,12 +11,18 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+import os
 from collections import Sequence
+from unittest import mock
 
 import pytest
 import torch
-from torch.utils.data import TensorDataset
+from torch.utils.data import DataLoader, TensorDataset
+from torch.utils.data.dataset import Dataset
+from torch.utils.data.distributed import DistributedSampler
+from torch.utils.data.sampler import Sampler
 
+from pytorch_lightning import Trainer
 from pytorch_lightning.trainer.supporters import (
     _nested_calc_num_data,
     CombinedDataset,
@@ -25,6 +31,7 @@
     CycleIterator,
     TensorRunningAccum,
 )
+from pytorch_lightning.utilities.apply_func import apply_to_collection
 from pytorch_lightning.utilities.exceptions import MisconfigurationException
 
 
@@ -237,3 +244,46 @@ def test_nested_calc_num_data(input_data, compute_func, expected_length):
     calculated_length = _nested_calc_num_data(input_data, compute_func)
 
     assert calculated_length == expected_length
+
+
+@mock.patch.dict(os.environ, {"CUDA_VISIBLE_DEVICES": "0,1", "PL_TRAINER_GPUS": "2"})
+@mock.patch('torch.cuda.device_count', return_value=2)
+@mock.patch('torch.cuda.is_available', return_value=True)
+def test_combined_data_loader_validation_test(cuda_available_mock, device_count_mock, tmpdir):
+    """
+    This test makes sure distributed sampler has been properly injected in dataloaders
+    when using CombinedLoader
+    """
+
+    class CustomDataset(Dataset):
+
+        def __init__(self, data):
+            self.data = data
+
+        def __len__(self):
+            return len(self.data)
+
+        def __getitem__(self, index):
+            return self.data[index]
+
+    dataloader = CombinedLoader({
+        "a": DataLoader(CustomDataset(range(10))),
+        "b": {
+            "c": DataLoader(CustomDataset(range(10))),
+            "d": DataLoader(CustomDataset(range(10)))
+        },
+        "e": [DataLoader(CustomDataset(range(10))),
+              DataLoader(CustomDataset(range(10)))]
+    })
+
+    trainer = Trainer(replace_sampler_ddp=True, accelerator="ddp", gpus=2)
+    dataloader = trainer.auto_add_sampler(dataloader, shuffle=True)
+    _count = 0
+
+    def _assert_distributed_sampler(v):
+        nonlocal _count
+        _count += 1
+        assert isinstance(v, DistributedSampler)
+
+    apply_to_collection(dataloader.sampler, Sampler, _assert_distributed_sampler)
+    assert _count == 5