fix

mmaction/models/backbones/resnet.py

@@ -554,7 +554,7 @@ def _load_torchvision_checkpoint(self, logger=None):
                 f': {remaining_names}')
 
     def init_weights(self):
-        if self.torchvision_pretrain:
+        if self.torchvision_pretrain and isinstance(self.pretrained, str):
             logger = get_root_logger()
             self._load_torchvision_checkpoint(logger)
         else:

mmaction/models/backbones/resnet3d.py

@@ -806,9 +806,7 @@ def init_weights(self, pretrained=None):
             self.init_cfg = dict(type='Pretrained', checkpoint=pretrained)
         if not self.pretrained2d:
             super().init_weights()
-        else:
-            assert pretrained is not None
-            self.pretrained = pretrained
+        elif isinstance(self.pretrained, str):
             logger = get_root_logger()
             self.inflate_weights(logger)
 

mmaction/models/backbones/resnet_tsm.py

@@ -1,13 +1,14 @@
 import torch
 import torch.nn as nn
 from mmcv.cnn import NonLocal3d
+from mmcv.runner import BaseModule
 from torch.nn.modules.utils import _ntuple
 
 from ..registry import BACKBONES
 from .resnet import ResNet
 
 
-class NL3DWrapper(nn.Module):
+class NL3DWrapper(BaseModule):
     """3D Non-local wrapper for ResNet50.
 
     Wrap ResNet layers with 3D NonLocal modules.
@@ -18,8 +19,12 @@ class NL3DWrapper(nn.Module):
         non_local_cfg (dict): Config for non-local layers. Default: ``dict()``.
     """
 
-    def __init__(self, block, num_segments, non_local_cfg=dict()):
-        super(NL3DWrapper, self).__init__()
+    def __init__(self,
+                 block,
+                 num_segments,
+                 non_local_cfg=dict(),
+                 init_cfg=None):
+        super().__init__(init_cfg)
         self.block = block
         self.non_local_cfg = non_local_cfg
         self.non_local_block = NonLocal3d(self.block.conv3.norm.num_features,

mmaction/models/backbones/x3d.py

@@ -3,7 +3,7 @@
 import torch.nn as nn
 import torch.utils.checkpoint as cp
 from mmcv.cnn import ConvModule, Swish, build_activation_layer
-from mmcv.runner import BaseModule
+from mmcv.runner import BaseModule, Sequential
 from mmcv.utils import _BatchNorm
 
 from ..registry import BACKBONES
@@ -230,6 +230,8 @@ def __init__(self,
         self.gamma_b = gamma_b
         self.gamma_d = gamma_d
 
+        self.zero_init_residual = zero_init_residual
+
         block_init_cfg = None
         assert not (init_cfg
                     and pretrained), ('init_cfg and pretrained cannot '
@@ -249,7 +251,7 @@ def __init__(self,
                 ]
                 if self.zero_init_residual:
                     block_init_cfg = dict(
-                        type='Constant', val=0, override=dict(name='norm3'))
+                        type='Constant', val=0, override=dict(name='conv3.bn'))
         else:
             raise TypeError('pretrained must be a str or None')
 
@@ -283,7 +285,6 @@ def __init__(self,
         self.act_cfg = act_cfg
         self.norm_eval = norm_eval
         self.with_cp = with_cp
-        self.zero_init_residual = zero_init_residual
 
         self.block = BlockX3D
         self.stage_blocks = self.stage_blocks[:num_stages]
@@ -451,7 +452,8 @@ def make_res_layer(self,
                     with_cp=with_cp,
                     **kwargs))
 
-        return nn.Sequential(*layers)
+        # return nn.Sequential(*layers)
+        return Sequential(*layers)
 
     def _make_stem_layer(self):
         """Construct the stem layers consists of a conv+norm+act module and a

mmaction/models/common/conv2plus1d.py

@@ -38,9 +38,9 @@ def __init__(self,
 
         if init_cfg is None:
             self.init_cfg = [
-                dict(type='kaiming', layer='Conv3d'),
+                dict(type='Kaiming', layer='Conv3d'),
                 dict(
-                    type='constant', val=1, layer=['_BatchNorm', 'GroupNorm'])
+                    type='Constant', val=1, layer=['_BatchNorm', 'GroupNorm'])
             ]
 
         kernel_size = _triple(kernel_size)

mmaction/models/common/conv_audio.py

@@ -41,9 +41,9 @@ def __init__(self,
 
         if init_cfg is None:
             self.init_cfg = [
-                dict(type='kaiming', layer='Conv3d'),
+                dict(type='Kaiming', layer='Conv3d'),
                 dict(
-                    type='constant', val=1, layer=['_BatchNorm', 'GroupNorm'])
+                    type='Constant', val=1, layer=['_BatchNorm', 'GroupNorm'])
             ]
 
         kernel_size = _pair(kernel_size)