[Enhance] Use PyTorch HSwish implementation. (#1709)

* [Enhance] Use PyTorch HSwish implementation.

* fix conv test

* upgrade version

* add version comments

mmcv/cnn/bricks/hswish.py

 # Copyright (c) OpenMMLab. All rights reserved.
 import torch.nn as nn
 
+from mmcv.utils import TORCH_VERSION, digit_version
 from .registry import ACTIVATION_LAYERS
 
 
-@ACTIVATION_LAYERS.register_module()
 class HSwish(nn.Module):
     """Hard Swish Module.
 
@@ -27,3 +27,12 @@ class HSwish(nn.Module):
 
     def forward(self, x):
         return x * self.act(x + 3) / 6
+
+
+if (TORCH_VERSION == 'parrots'
+        or digit_version(TORCH_VERSION) < digit_version('1.7')):
+    # Hardswish is not supported when PyTorch version < 1.6.
+    # And Hardswish in PyTorch 1.6 does not support inplace.
+    ACTIVATION_LAYERS.register_module(module=HSwish)
+else:
+    ACTIVATION_LAYERS.register_module(module=nn.Hardswish, name='HSwish')

tests/test_cnn/test_conv_module.py

@@ -6,6 +6,7 @@ import torch
 import torch.nn as nn
 
 from mmcv.cnn.bricks import CONV_LAYERS, ConvModule, HSigmoid, HSwish
+from mmcv.utils import TORCH_VERSION, digit_version
 
 
 @CONV_LAYERS.register_module()
@@ -138,7 +139,12 @@ def test_conv_module():
 
     # HSwish
     conv = ConvModule(3, 8, 3, padding=1, act_cfg=dict(type='HSwish'))
-    assert isinstance(conv.activate, HSwish)
+    if (TORCH_VERSION == 'parrots'
+            or digit_version(TORCH_VERSION) < digit_version('1.7')):
+        assert isinstance(conv.activate, HSwish)
+    else:
+        assert isinstance(conv.activate, nn.Hardswish)
+
     output = conv(x)
     assert output.shape == (1, 8, 256, 256)