Replace convert_syncbatchnorm in mmseg (#93)

* replace convert_syncbatchnorm with revert_sync_batchnorm from mmcv

* change logger

Author: RunningLeon

configs/mmseg/segmentation_pplnn_static-512x1024.py

@@ -0,0 +1,5 @@
+_base_ = ['./segmentation_static.py', '../_base_/backends/pplnn.py']
+
+onnx_config = dict(input_shape=[1024, 512])
+
+backend_config = dict(model_inputs=dict(opt_shape=[1, 3, 512, 1024]))

docs/en/codebases/mmseg.md

@@ -15,7 +15,7 @@ Please refer to [get_started.md](https://github.com/open-mmlab/mmsegmentation/bl
 | DeepLabV3                   |      Y      |    Y     |  Y   |   Y   |    Y     |   [config](https://github.com/open-mmlab/mmsegmentation/tree/master/configs/deeplabv3)   |
 | DeepLabV3+                  |      Y      |    Y     |  Y   |   Y   |    Y     | [config](https://github.com/open-mmlab/mmsegmentation/tree/master/configs/deeplabv3plus) |
 | Fast-SCNN[*](#static_shape) |      Y      |    Y     |  N   |   Y   |    Y     |   [config](https://github.com/open-mmlab/mmsegmentation/tree/master/configs/fastscnn)    |
-| UNet                        |      Y      |    Y     |  Y   |   N   |    Y     |     [config](https://github.com/open-mmlab/mmsegmentation/tree/master/configs/unet)      |
+| UNet[*](#static_shape)      |      Y      |    Y     |  Y   |   Y   |    Y     |     [config](https://github.com/open-mmlab/mmsegmentation/tree/master/configs/unet)      |
 
 ### Reminder
 

docs/en/supported_models.md

@@ -28,7 +28,7 @@ The table below lists the models that are guaranteed to be exportable to other b
 | DeepLabV3                 | MMSegmentation   |      Y      |    Y     |  Y   |   Y   |    Y     |      [config](https://github.com/open-mmlab/mmsegmentation/tree/master/configs/deeplabv3)      |
 | DeepLabV3+                | MMSegmentation   |      Y      |    Y     |  Y   |   Y   |    Y     |    [config](https://github.com/open-mmlab/mmsegmentation/tree/master/configs/deeplabv3plus)    |
 | Fast-SCNN[*static](#note) | MMSegmentation   |      Y      |    Y     |  N   |   Y   |    Y     |      [config](https://github.com/open-mmlab/mmsegmentation/tree/master/configs/fastscnn)       |
-| UNet                      | MMSegmentation   |      Y      |    Y     |  Y   |   N   |    Y     |        [config](https://github.com/open-mmlab/mmsegmentation/tree/master/configs/unet)         |
+| UNet[*static](#note)      | MMSegmentation   |      Y      |    Y     |  Y   |   Y   |    Y     |        [config](https://github.com/open-mmlab/mmsegmentation/tree/master/configs/unet)         |
 | SRCNN                     | MMEditing        |      Y      |    Y     |  Y   |   Y   |    Y     |     [config](https://github.com/open-mmlab/mmediting/tree/master/configs/restorers/srcnn)      |
 | ESRGAN                    | MMEditing        |      Y      |    Y     |  Y   |   Y   |    Y     |     [config](https://github.com/open-mmlab/mmediting/tree/master/configs/restorers/esrgan)     |
 | SRGAN                     | MMEditing        |      Y      |    Y     |  Y   |   Y   |    Y     | [config](https://github.com/open-mmlab/mmediting/tree/master/configs/restorers/srresnet_srgan) |

mmdeploy/codebase/mmseg/deploy/__init__.py

@@ -1,6 +1,5 @@
 # Copyright (c) OpenMMLab. All rights reserved.
 from .mmsegmentation import MMSegmentation
 from .segmentation import Segmentation
-from .utils import convert_syncbatchnorm
 
-__all__ = ['convert_syncbatchnorm', 'MMSegmentation', 'Segmentation']
+__all__ = ['MMSegmentation', 'Segmentation']

mmdeploy/codebase/mmseg/deploy/segmentation.py

@@ -89,10 +89,9 @@ def init_pytorch_model(self,
                 codebases.
         """
         from mmseg.apis import init_segmentor
-        from mmdeploy.codebase.mmseg.deploy import convert_syncbatchnorm
+        from mmcv.cnn.utils import revert_sync_batchnorm
         model = init_segmentor(self.model_cfg, model_checkpoint, self.device)
-        model = convert_syncbatchnorm(model)
-
+        model = revert_sync_batchnorm(model)
         return model.eval()
 
     def create_input(self,