[Fix] Remove unit test temporarily

tests/test_model/test_wrappers/test_model_wrapper.py

@@ -11,7 +11,7 @@
 from mmengine.dist import all_gather, broadcast
 from mmengine.model import (BaseDataPreprocessor, BaseModel,
                             ExponentialMovingAverage,
-                            MMDistributedDataParallel, MMPipelineParallel,
+                            MMDistributedDataParallel,
                             MMSeparateDistributedDataParallel)
 from mmengine.optim import AmpOptimWrapper, OptimWrapper, OptimWrapperDict
 from mmengine.testing import assert_allclose
@@ -48,24 +48,6 @@ def forward(self, inputs, data_sample=None, mode='tensor'):
             return x
 
 
-class ToyLinearModel(BaseModel):
-    # because the flop analyzer cannot analyze the conv layer
-    def __init__(self):
-        super().__init__(data_preprocessor=ToyDataPreprocessor())
-        self.linear1 = nn.Linear(1, 1)
-        self.linear2 = nn.Linear(1, 2)
-
-    def forward(self, inputs, data_sample=None, mode='tensor'):
-        x = self.linear1(inputs)
-        x = self.linear2(x)
-        if mode == 'loss':
-            return dict(loss=x)
-        elif mode == 'predict':
-            return x
-        else:
-            return x
-
-
 class ComplexModel(BaseModel):
 
     def __init__(self):
@@ -316,36 +298,3 @@ def test_test_step(self):
         data = dict(inputs=inputs, data_sample=MagicMock())
         predictions = fsdp_model.test_step(data)
         self.assertIsInstance(predictions, torch.Tensor)
-
-
-@unittest.skipIf(
-    torch.cuda.device_count() < 2,
-    reason='need 2 or more gpu to test pipeline parallel')
-@unittest.skipIf(
-    digit_version(TORCH_VERSION) < digit_version('2.0.0'),
-    reason='pipeline parallelism needs Pytorch 2.0.0 or higher')
-class TestMMPipelineParallel(unittest.TestCase):
-
-    def test_init(self):
-        model = ToyLinearModel()
-        model = MMPipelineParallel(model, num_pipelines=2)
-        self.assertEqual(model.model.linear1.weight.device,
-                         torch.device('meta'))
-        self.assertEqual(model.model.linear2.weight.device,
-                         torch.device('meta'))
-
-    def test_forward(self):
-        model = ToyLinearModel()
-        model = MMPipelineParallel(model, num_pipelines=2)
-        inputs = torch.randn(6, 1)
-        self.assertIsInstance(model(inputs), torch.Tensor)
-        self.assertEqual(model.model.linear1.weight.device,
-                         torch.device('cuda:0'))
-        self.assertEqual(model.model.linear2.weight.device,
-                         torch.device('cuda:1'))
-        self.assertEqual(model(inputs).shape, torch.Size([6, 2]))
-        self.assertEqual(model(inputs).device, torch.device('cuda:1'))
-        self.assertIn('linear1', model.device_map)
-        self.assertIn('linear1', model.device_map)
-        self.assertEqual(model.device_map['linear1']['part_id'], 0)
-        self.assertEqual(model.device_map['linear2']['part_id'], 1)