[Doc]:add type hints for res_layer, se_layer,normed_predictor,positio…

…nal_encoding (#9346)

* [Fix]:add type hints for res_layer

* Update mmdet/models/layers/res_layer.py

Co-authored-by: BigDong <[email protected]>

* [Fix]:add type hint of norm1 and norm2

* [WIP]:add res_layer type hints

* [WIP]:add layer type hints about issue 9234

* [FIX]:add all type hints and change some function note

* [FIX]:add docstrings and change default -> defaults

Co-authored-by: BigDong <[email protected]>

mmdet/models/layers/normed_predictor.py

@@ -2,6 +2,7 @@
 import torch
 import torch.nn as nn
 import torch.nn.functional as F
+from torch import Tensor
 
 from mmdet.registry import MODELS
 
@@ -13,25 +14,32 @@ class NormedLinear(nn.Linear):
     """Normalized Linear Layer.
 
     Args:
-        tempeature (float, optional): Tempeature term. Default to 20.
-        power (int, optional): Power term. Default to 1.0.
+        tempeature (float, optional): Tempeature term. Defaults to 20.
+        power (int, optional): Power term. Defaults to 1.0.
         eps (float, optional): The minimal value of divisor to
-             keep numerical stability. Default to 1e-6.
+             keep numerical stability. Defaults to 1e-6.
     """
 
-    def __init__(self, *args, tempearture=20, power=1.0, eps=1e-6, **kwargs):
-        super(NormedLinear, self).__init__(*args, **kwargs)
+    def __init__(self,
+                 *args,
+                 tempearture: float = 20,
+                 power: int = 1.0,
+                 eps: float = 1e-6,
+                 **kwargs) -> None:
+        super().__init__(*args, **kwargs)
         self.tempearture = tempearture
         self.power = power
         self.eps = eps
         self.init_weights()
 
-    def init_weights(self):
+    def init_weights(self) -> None:
+        """Initialize the weights."""
         nn.init.normal_(self.weight, mean=0, std=0.01)
         if self.bias is not None:
             nn.init.constant_(self.bias, 0)
 
-    def forward(self, x):
+    def forward(self, x: Tensor) -> Tensor:
+        """Forward function for `NormedLinear`."""
         weight_ = self.weight / (
             self.weight.norm(dim=1, keepdim=True).pow(self.power) + self.eps)
         x_ = x / (x.norm(dim=1, keepdim=True).pow(self.power) + self.eps)
@@ -45,28 +53,29 @@ class NormedConv2d(nn.Conv2d):
     """Normalized Conv2d Layer.
 
     Args:
-        tempeature (float, optional): Tempeature term. Default to 20.
-        power (int, optional): Power term. Default to 1.0.
+        tempeature (float, optional): Tempeature term. Defaults to 20.
+        power (int, optional): Power term. Defaults to 1.0.
         eps (float, optional): The minimal value of divisor to
-             keep numerical stability. Default to 1e-6.
+             keep numerical stability. Defaults to 1e-6.
         norm_over_kernel (bool, optional): Normalize over kernel.
-             Default to False.
+             Defaults to False.
     """
 
     def __init__(self,
                  *args,
-                 tempearture=20,
-                 power=1.0,
-                 eps=1e-6,
-                 norm_over_kernel=False,
-                 **kwargs):
-        super(NormedConv2d, self).__init__(*args, **kwargs)
+                 tempearture: float = 20,
+                 power: int = 1.0,
+                 eps: float = 1e-6,
+                 norm_over_kernel: bool = False,
+                 **kwargs) -> None:
+        super().__init__(*args, **kwargs)
         self.tempearture = tempearture
         self.power = power
         self.norm_over_kernel = norm_over_kernel
         self.eps = eps
 
-    def forward(self, x):
+    def forward(self, x: Tensor) -> Tensor:
+        """Forward function for `NormedConv2d`."""
         if not self.norm_over_kernel:
             weight_ = self.weight / (
                 self.weight.norm(dim=1, keepdim=True).pow(self.power) +

mmdet/models/layers/positional_encoding.py

@@ -4,8 +4,10 @@
 import torch
 import torch.nn as nn
 from mmengine.model import BaseModule
+from torch import Tensor
 
 from mmdet.registry import MODELS
+from mmdet.utils import MultiConfig, OptMultiConfig
 
 
 @MODELS.register_module()
@@ -31,18 +33,18 @@ class SinePositionalEncoding(BaseModule):
         offset (float): offset add to embed when do the normalization.
             Defaults to 0.
         init_cfg (dict or list[dict], optional): Initialization config dict.
-            Default: None
+            Defaults to None
     """
 
     def __init__(self,
-                 num_feats,
-                 temperature=10000,
-                 normalize=False,
-                 scale=2 * math.pi,
-                 eps=1e-6,
-                 offset=0.,
-                 init_cfg=None):
-        super(SinePositionalEncoding, self).__init__(init_cfg)
+                 num_feats: int,
+                 temperature: int = 10000,
+                 normalize: bool = False,
+                 scale: float = 2 * math.pi,
+                 eps: float = 1e-6,
+                 offset: float = 0.,
+                 init_cfg: OptMultiConfig = None) -> None:
+        super().__init__(init_cfg=init_cfg)
         if normalize:
             assert isinstance(scale, (float, int)), 'when normalize is set,' \
                 'scale should be provided and in float or int type, ' \
@@ -54,7 +56,7 @@ def __init__(self,
         self.eps = eps
         self.offset = offset
 
-    def forward(self, mask):
+    def forward(self, mask: Tensor) -> Tensor:
         """Forward function for `SinePositionalEncoding`.
 
         Args:
@@ -93,7 +95,7 @@ def forward(self, mask):
         pos = torch.cat((pos_y, pos_x), dim=3).permute(0, 3, 1, 2)
         return pos
 
-    def __repr__(self):
+    def __repr__(self) -> str:
         """str: a string that describes the module"""
         repr_str = self.__class__.__name__
         repr_str += f'(num_feats={self.num_feats}, '
@@ -113,25 +115,27 @@ class LearnedPositionalEncoding(BaseModule):
             along x-axis or y-axis. The final returned dimension for
             each position is 2 times of this value.
         row_num_embed (int, optional): The dictionary size of row embeddings.
-            Default 50.
+            Defaults to 50.
         col_num_embed (int, optional): The dictionary size of col embeddings.
-            Default 50.
+            Defaults to 50.
         init_cfg (dict or list[dict], optional): Initialization config dict.
     """
 
-    def __init__(self,
-                 num_feats,
-                 row_num_embed=50,
-                 col_num_embed=50,
-                 init_cfg=dict(type='Uniform', layer='Embedding')):
-        super(LearnedPositionalEncoding, self).__init__(init_cfg)
+    def __init__(
+        self,
+        num_feats: int,
+        row_num_embed: int = 50,
+        col_num_embed: int = 50,
+        init_cfg: MultiConfig = dict(type='Uniform', layer='Embedding')
+    ) -> None:
+        super().__init__(init_cfg=init_cfg)
         self.row_embed = nn.Embedding(row_num_embed, num_feats)
         self.col_embed = nn.Embedding(col_num_embed, num_feats)
         self.num_feats = num_feats
         self.row_num_embed = row_num_embed
         self.col_num_embed = col_num_embed
 
-    def forward(self, mask):
+    def forward(self, mask: Tensor) -> Tensor:
         """Forward function for `LearnedPositionalEncoding`.
 
         Args:
@@ -155,7 +159,7 @@ def forward(self, mask):
                             1).unsqueeze(0).repeat(mask.shape[0], 1, 1, 1)
         return pos
 
-    def __repr__(self):
+    def __repr__(self) -> str:
         """str: a string that describes the module"""
         repr_str = self.__class__.__name__
         repr_str += f'(num_feats={self.num_feats}, '

mmdet/models/layers/res_layer.py

@@ -1,8 +1,13 @@
 # Copyright (c) OpenMMLab. All rights reserved.
+from typing import Optional
+
 from mmcv.cnn import build_conv_layer, build_norm_layer
 from mmengine.model import BaseModule, Sequential
+from torch import Tensor
 from torch import nn as nn
 
+from mmdet.utils import ConfigType, OptConfigType, OptMultiConfig
+
 
 class ResLayer(Sequential):
     """ResLayer to build ResNet style backbone.
@@ -12,28 +17,28 @@ class ResLayer(Sequential):
         inplanes (int): inplanes of block.
         planes (int): planes of block.
         num_blocks (int): number of blocks.
-        stride (int): stride of the first block. Default: 1
+        stride (int): stride of the first block. Defaults to 1
         avg_down (bool): Use AvgPool instead of stride conv when
-            downsampling in the bottleneck. Default: False
+            downsampling in the bottleneck. Defaults to False
         conv_cfg (dict): dictionary to construct and config conv layer.
-            Default: None
+            Defaults to None
         norm_cfg (dict): dictionary to construct and config norm layer.
-            Default: dict(type='BN')
+            Defaults to dict(type='BN')
         downsample_first (bool): Downsample at the first block or last block.
-            False for Hourglass, True for ResNet. Default: True
+            False for Hourglass, True for ResNet. Defaults to True
     """
 
     def __init__(self,
-                 block,
-                 inplanes,
-                 planes,
-                 num_blocks,
-                 stride=1,
-                 avg_down=False,
-                 conv_cfg=None,
-                 norm_cfg=dict(type='BN'),
-                 downsample_first=True,
-                 **kwargs):
+                 block: BaseModule,
+                 inplanes: int,
+                 planes: int,
+                 num_blocks: int,
+                 stride: int = 1,
+                 avg_down: bool = False,
+                 conv_cfg: OptConfigType = None,
+                 norm_cfg: ConfigType = dict(type='BN'),
+                 downsample_first: bool = True,
+                 **kwargs) -> None:
         self.block = block
 
         downsample = None
@@ -101,7 +106,7 @@ def __init__(self,
                     conv_cfg=conv_cfg,
                     norm_cfg=norm_cfg,
                     **kwargs))
-        super(ResLayer, self).__init__(*layers)
+        super().__init__(*layers)
 
 
 class SimplifiedBasicBlock(BaseModule):
@@ -114,19 +119,19 @@ class SimplifiedBasicBlock(BaseModule):
     expansion = 1
 
     def __init__(self,
-                 inplanes,
-                 planes,
-                 stride=1,
-                 dilation=1,
-                 downsample=None,
-                 style='pytorch',
-                 with_cp=False,
-                 conv_cfg=None,
-                 norm_cfg=dict(type='BN'),
-                 dcn=None,
-                 plugins=None,
-                 init_fg=None):
-        super(SimplifiedBasicBlock, self).__init__(init_fg)
+                 inplanes: int,
+                 planes: int,
+                 stride: int = 1,
+                 dilation: int = 1,
+                 downsample: Optional[Sequential] = None,
+                 style: ConfigType = 'pytorch',
+                 with_cp: bool = False,
+                 conv_cfg: OptConfigType = None,
+                 norm_cfg: ConfigType = dict(type='BN'),
+                 dcn: OptConfigType = None,
+                 plugins: OptConfigType = None,
+                 init_cfg: OptMultiConfig = None) -> None:
+        super().__init__(init_cfg=init_cfg)
         assert dcn is None, 'Not implemented yet.'
         assert plugins is None, 'Not implemented yet.'
         assert not with_cp, 'Not implemented yet.'
@@ -159,17 +164,17 @@ def __init__(self,
         self.with_cp = with_cp
 
     @property
-    def norm1(self):
+    def norm1(self) -> Optional[BaseModule]:
         """nn.Module: normalization layer after the first convolution layer"""
         return getattr(self, self.norm1_name) if self.with_norm else None
 
     @property
-    def norm2(self):
+    def norm2(self) -> Optional[BaseModule]:
         """nn.Module: normalization layer after the second convolution layer"""
         return getattr(self, self.norm2_name) if self.with_norm else None
 
-    def forward(self, x):
-        """Forward function."""
+    def forward(self, x: Tensor) -> Tensor:
+        """Forward function for SimplifiedBasicBlock."""
 
         identity = x
 

mmdet/models/layers/se_layer.py

@@ -15,17 +15,17 @@ class SELayer(BaseModule):
     Args:
         channels (int): The input (and output) channels of the SE layer.
         ratio (int): Squeeze ratio in SELayer, the intermediate channel will be
-            ``int(channels/ratio)``. Default: 16.
+            ``int(channels/ratio)``. Defaults to 16.
         conv_cfg (None or dict): Config dict for convolution layer.
-            Default: None, which means using conv2d.
+            Defaults to None, which means using conv2d.
         act_cfg (dict or Sequence[dict]): Config dict for activation layer.
             If act_cfg is a dict, two activation layers will be configurated
             by this dict. If act_cfg is a sequence of dicts, the first
             activation layer will be configurated by the first dict and the
             second activation layer will be configurated by the second dict.
-            Default: (dict(type='ReLU'), dict(type='Sigmoid'))
+            Defaults to (dict(type='ReLU'), dict(type='Sigmoid'))
         init_cfg (dict or list[dict], optional): Initialization config dict.
-            Default: None
+            Defaults to None
     """
 
     def __init__(self,
@@ -57,6 +57,7 @@ def __init__(self,
             act_cfg=act_cfg[1])
 
     def forward(self, x: Tensor) -> Tensor:
+        """Forward function for SELayer."""
         out = self.global_avgpool(x)
         out = self.conv1(out)
         out = self.conv2(out)
@@ -75,18 +76,18 @@ class DyReLU(BaseModule):
         channels (int): The input (and output) channels of DyReLU module.
         ratio (int): Squeeze ratio in Squeeze-and-Excitation-like module,
             the intermediate channel will be ``int(channels/ratio)``.
-            Default: 4.
+            Defaults to 4.
         conv_cfg (None or dict): Config dict for convolution layer.
-            Default: None, which means using conv2d.
+            Defaults to None, which means using conv2d.
         act_cfg (dict or Sequence[dict]): Config dict for activation layer.
             If act_cfg is a dict, two activation layers will be configurated
             by this dict. If act_cfg is a sequence of dicts, the first
             activation layer will be configurated by the first dict and the
             second activation layer will be configurated by the second dict.
-            Default: (dict(type='ReLU'), dict(type='HSigmoid', bias=3.0,
+            Defaults to (dict(type='ReLU'), dict(type='HSigmoid', bias=3.0,
             divisor=6.0))
         init_cfg (dict or list[dict], optional): Initialization config dict.
-            Default: None
+            Defaults to None
     """
 
     def __init__(self,
@@ -140,7 +141,7 @@ class ChannelAttention(BaseModule):
     Args:
         channels (int): The input (and output) channels of the attention layer.
         init_cfg (dict or list[dict], optional): Initialization config dict.
-            Defaults to None.
+            Defaults to None
     """
 
     def __init__(self, channels: int, init_cfg: OptMultiConfig = None) -> None:
@@ -150,6 +151,7 @@ def __init__(self, channels: int, init_cfg: OptMultiConfig = None) -> None:
         self.act = nn.Hardsigmoid(inplace=True)
 
     def forward(self, x: Tensor) -> Tensor:
+        """Forward function for ChannelAttention."""
         out = self.global_avgpool(x)
         out = self.fc(out)
         out = self.act(out)