File size: 12,628 Bytes
dfca1b5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
import warnings
from typing import Callable, List, Optional

import torch
from torch import Tensor

interpolate = torch.nn.functional.interpolate


class FrozenBatchNorm2d(torch.nn.Module):
    """
    BatchNorm2d where the batch statistics and the affine parameters are fixed

    Args:
        num_features (int): Number of features ``C`` from an expected input of size ``(N, C, H, W)``
        eps (float): a value added to the denominator for numerical stability. Default: 1e-5
    """

    def __init__(
        self,
        num_features: int,
        eps: float = 1e-5,
    ):
        super().__init__()
        # _log_api_usage_once(self)
        self.eps = eps
        self.register_buffer("weight", torch.ones(num_features))
        self.register_buffer("bias", torch.zeros(num_features))
        self.register_buffer("running_mean", torch.zeros(num_features))
        self.register_buffer("running_var", torch.ones(num_features))

    def _load_from_state_dict(
        self,
        state_dict: dict,
        prefix: str,
        local_metadata: dict,
        strict: bool,
        missing_keys: List[str],
        unexpected_keys: List[str],
        error_msgs: List[str],
    ):
        num_batches_tracked_key = prefix + "num_batches_tracked"
        if num_batches_tracked_key in state_dict:
            del state_dict[num_batches_tracked_key]

        super()._load_from_state_dict(
            state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs
        )

    def forward(self, x: Tensor) -> Tensor:
        # move reshapes to the beginning
        # to make it fuser-friendly
        w = self.weight.reshape(1, -1, 1, 1)
        b = self.bias.reshape(1, -1, 1, 1)
        rv = self.running_var.reshape(1, -1, 1, 1)
        rm = self.running_mean.reshape(1, -1, 1, 1)
        scale = w * (rv + self.eps).rsqrt()
        bias = b - rm * scale
        return x * scale + bias

    def __repr__(self) -> str:
        return f"{self.__class__.__name__}({self.weight.shape[0]}, eps={self.eps})"


class ConvNormActivation(torch.nn.Sequential):
    def __init__(
        self,
        in_channels: int,
        out_channels: int,
        kernel_size: int = 3,
        stride: int = 1,
        padding: Optional[int] = None,
        groups: int = 1,
        norm_layer: Optional[Callable[..., torch.nn.Module]] = torch.nn.BatchNorm2d,
        activation_layer: Optional[Callable[..., torch.nn.Module]] = torch.nn.ReLU,
        dilation: int = 1,
        inplace: Optional[bool] = True,
        bias: Optional[bool] = None,
        conv_layer: Callable[..., torch.nn.Module] = torch.nn.Conv2d,
    ) -> None:

        if padding is None:
            padding = (kernel_size - 1) // 2 * dilation
        if bias is None:
            bias = norm_layer is None

        layers = [
            conv_layer(
                in_channels,
                out_channels,
                kernel_size,
                stride,
                padding,
                dilation=dilation,
                groups=groups,
                bias=bias,
            )
        ]

        if norm_layer is not None:
            layers.append(norm_layer(out_channels))

        if activation_layer is not None:
            params = {} if inplace is None else {"inplace": inplace}
            layers.append(activation_layer(**params))
        super().__init__(*layers)
        # _log_api_usage_once(self)
        self.out_channels = out_channels

        if self.__class__ == ConvNormActivation:
            warnings.warn(
                "Don't use ConvNormActivation directly, please use Conv2dNormActivation and Conv3dNormActivation instead."
            )


class Conv2dNormActivation(ConvNormActivation):
    """
    Configurable block used for Convolution2d-Normalization-Activation blocks.

    Args:
        in_channels (int): Number of channels in the input image
        out_channels (int): Number of channels produced by the Convolution-Normalization-Activation block
        kernel_size: (int, optional): Size of the convolving kernel. Default: 3
        stride (int, optional): Stride of the convolution. Default: 1
        padding (int, tuple or str, optional): Padding added to all four sides of the input. Default: None, in which case it will calculated as ``padding = (kernel_size - 1) // 2 * dilation``
        groups (int, optional): Number of blocked connections from input channels to output channels. Default: 1
        norm_layer (Callable[..., torch.nn.Module], optional): Norm layer that will be stacked on top of the convolution layer. If ``None`` this layer wont be used. Default: ``torch.nn.BatchNorm2d``
        activation_layer (Callable[..., torch.nn.Module], optional): Activation function which will be stacked on top of the normalization layer (if not None), otherwise on top of the conv layer. If ``None`` this layer wont be used. Default: ``torch.nn.ReLU``
        dilation (int): Spacing between kernel elements. Default: 1
        inplace (bool): Parameter for the activation layer, which can optionally do the operation in-place. Default ``True``
        bias (bool, optional): Whether to use bias in the convolution layer. By default, biases are included if ``norm_layer is None``.

    """

    def __init__(
        self,
        in_channels: int,
        out_channels: int,
        kernel_size: int = 3,
        stride: int = 1,
        padding: Optional[int] = None,
        groups: int = 1,
        norm_layer: Optional[Callable[..., torch.nn.Module]] = torch.nn.BatchNorm2d,
        activation_layer: Optional[Callable[..., torch.nn.Module]] = torch.nn.ReLU,
        dilation: int = 1,
        inplace: Optional[bool] = True,
        bias: Optional[bool] = None,
    ) -> None:

        super().__init__(
            in_channels,
            out_channels,
            kernel_size,
            stride,
            padding,
            groups,
            norm_layer,
            activation_layer,
            dilation,
            inplace,
            bias,
            torch.nn.Conv2d,
        )


class Conv3dNormActivation(ConvNormActivation):
    """
    Configurable block used for Convolution3d-Normalization-Activation blocks.

    Args:
        in_channels (int): Number of channels in the input video.
        out_channels (int): Number of channels produced by the Convolution-Normalization-Activation block
        kernel_size: (int, optional): Size of the convolving kernel. Default: 3
        stride (int, optional): Stride of the convolution. Default: 1
        padding (int, tuple or str, optional): Padding added to all four sides of the input. Default: None, in which case it will calculated as ``padding = (kernel_size - 1) // 2 * dilation``
        groups (int, optional): Number of blocked connections from input channels to output channels. Default: 1
        norm_layer (Callable[..., torch.nn.Module], optional): Norm layer that will be stacked on top of the convolution layer. If ``None`` this layer wont be used. Default: ``torch.nn.BatchNorm3d``
        activation_layer (Callable[..., torch.nn.Module], optional): Activation function which will be stacked on top of the normalization layer (if not None), otherwise on top of the conv layer. If ``None`` this layer wont be used. Default: ``torch.nn.ReLU``
        dilation (int): Spacing between kernel elements. Default: 1
        inplace (bool): Parameter for the activation layer, which can optionally do the operation in-place. Default ``True``
        bias (bool, optional): Whether to use bias in the convolution layer. By default, biases are included if ``norm_layer is None``.
    """

    def __init__(
        self,
        in_channels: int,
        out_channels: int,
        kernel_size: int = 3,
        stride: int = 1,
        padding: Optional[int] = None,
        groups: int = 1,
        norm_layer: Optional[Callable[..., torch.nn.Module]] = torch.nn.BatchNorm3d,
        activation_layer: Optional[Callable[..., torch.nn.Module]] = torch.nn.ReLU,
        dilation: int = 1,
        inplace: Optional[bool] = True,
        bias: Optional[bool] = None,
    ) -> None:

        super().__init__(
            in_channels,
            out_channels,
            kernel_size,
            stride,
            padding,
            groups,
            norm_layer,
            activation_layer,
            dilation,
            inplace,
            bias,
            torch.nn.Conv3d,
        )


class SqueezeExcitation(torch.nn.Module):
    """
    This block implements the Squeeze-and-Excitation block from https://arxiv.org/abs/1709.01507 (see Fig. 1).
    Parameters ``activation``, and ``scale_activation`` correspond to ``delta`` and ``sigma`` in eq. 3.

    Args:
        input_channels (int): Number of channels in the input image
        squeeze_channels (int): Number of squeeze channels
        activation (Callable[..., torch.nn.Module], optional): ``delta`` activation. Default: ``torch.nn.ReLU``
        scale_activation (Callable[..., torch.nn.Module]): ``sigma`` activation. Default: ``torch.nn.Sigmoid``
    """

    def __init__(
        self,
        input_channels: int,
        squeeze_channels: int,
        activation: Callable[..., torch.nn.Module] = torch.nn.ReLU,
        scale_activation: Callable[..., torch.nn.Module] = torch.nn.Sigmoid,
    ) -> None:
        super().__init__()
        # _log_api_usage_once(self)
        self.avgpool = torch.nn.AdaptiveAvgPool2d(1)
        self.fc1 = torch.nn.Conv2d(input_channels, squeeze_channels, 1)
        self.fc2 = torch.nn.Conv2d(squeeze_channels, input_channels, 1)
        self.activation = activation()
        self.scale_activation = scale_activation()

    def _scale(self, input: Tensor) -> Tensor:
        scale = self.avgpool(input)
        scale = self.fc1(scale)
        scale = self.activation(scale)
        scale = self.fc2(scale)
        return self.scale_activation(scale)

    def forward(self, input: Tensor) -> Tensor:
        scale = self._scale(input)
        return scale * input


class MLP(torch.nn.Sequential):
    """This block implements the multi-layer perceptron (MLP) module.

    Args:
        in_channels (int): Number of channels of the input
        hidden_channels (List[int]): List of the hidden channel dimensions
        norm_layer (Callable[..., torch.nn.Module], optional): Norm layer that will be stacked on top of the convolution layer. If ``None`` this layer wont be used. Default: ``None``
        activation_layer (Callable[..., torch.nn.Module], optional): Activation function which will be stacked on top of the normalization layer (if not None), otherwise on top of the conv layer. If ``None`` this layer wont be used. Default: ``torch.nn.ReLU``
        inplace (bool): Parameter for the activation layer, which can optionally do the operation in-place. Default ``True``
        bias (bool): Whether to use bias in the linear layer. Default ``True``
        dropout (float): The probability for the dropout layer. Default: 0.0
    """

    def __init__(
        self,
        in_channels: int,
        hidden_channels: List[int],
        norm_layer: Optional[Callable[..., torch.nn.Module]] = None,
        activation_layer: Optional[Callable[..., torch.nn.Module]] = torch.nn.ReLU,
        inplace: Optional[bool] = True,
        bias: bool = True,
        dropout: float = 0.0,
    ):
        # The addition of `norm_layer` is inspired from the implementation of TorchMultimodal:
        # https://github.com/facebookresearch/multimodal/blob/5dec8a/torchmultimodal/modules/layers/mlp.py
        params = {} if inplace is None else {"inplace": inplace}

        layers = []
        in_dim = in_channels
        for hidden_dim in hidden_channels[:-1]:
            layers.append(torch.nn.Linear(in_dim, hidden_dim, bias=bias))
            if norm_layer is not None:
                layers.append(norm_layer(hidden_dim))
            layers.append(activation_layer(**params))
            layers.append(torch.nn.Dropout(dropout, **params))
            in_dim = hidden_dim

        layers.append(torch.nn.Linear(in_dim, hidden_channels[-1], bias=bias))
        layers.append(torch.nn.Dropout(dropout, **params))

        super().__init__(*layers)
        # _log_api_usage_once(self)


class Permute(torch.nn.Module):
    """This module returns a view of the tensor input with its dimensions permuted.

    Args:
        dims (List[int]): The desired ordering of dimensions
    """

    def __init__(self, dims: List[int]):
        super().__init__()
        self.dims = dims

    def forward(self, x: Tensor) -> Tensor:
        return torch.permute(x, self.dims)