| import torch |
| import torch.nn as nn |
| import torch.nn.functional as F |
| from timm.models.layers import trunc_normal_, DropPath |
| from functools import partial |
| import numpy as np |
| from scipy.spatial import distance |
|
|
| class LayerNorm(nn.Module): |
| r""" LayerNorm that supports two data formats: channels_last (default) or channels_first. |
| The ordering of the dimensions in the inputs. channels_last corresponds to inputs with |
| shape (batch_size, height, width, channels) while channels_first corresponds to inputs |
| with shape (batch_size, channels, height, width). |
| """ |
| def __init__(self, normalized_shape, eps=1e-6, data_format="channels_last"): |
| super().__init__() |
| self.weight = nn.Parameter(torch.ones(normalized_shape)) |
| self.bias = nn.Parameter(torch.zeros(normalized_shape)) |
| self.eps = eps |
| self.data_format = data_format |
| if self.data_format not in ["channels_last", "channels_first"]: |
| raise NotImplementedError |
| self.normalized_shape = (normalized_shape, ) |
|
|
| def forward(self, x): |
| if self.data_format == "channels_last": |
| return F.layer_norm(x, self.normalized_shape, self.weight, self.bias, self.eps) |
| elif self.data_format == "channels_first": |
| u = x.mean(1, keepdim=True) |
| s = (x - u).pow(2).mean(1, keepdim=True) |
| x = (x - u) / torch.sqrt(s + self.eps) |
| |
| x = self.weight[:, None, None, None] * x + self.bias[:, None, None, None] |
|
|
| return x |
|
|
| class repux_block(nn.Module): |
| r""" ConvNeXt Block. There are two equivalent implementations: |
| (1) DwConv -> LayerNorm (channels_first) -> 1x1 Conv -> GELU -> 1x1 Conv; all in (N, C, H, W) |
| (2) DwConv -> Permute to (N, H, W, C); LayerNorm (channels_last) -> Linear -> GELU -> Linear; Permute back |
| We use (2) as we find it slightly faster in PyTorch |
| |
| Args: |
| dim (int): Number of input channels. |
| drop_path (float): Stochastic depth rate. Default: 0.0 |
| layer_scale_init_value (float): Init value for Layer Scale. Default: 1e-6. |
| """ |
|
|
| def __init__(self, dim, ks, a, drop_path=0., layer_scale_init_value=1e-6, deploy=False): |
| super().__init__() |
| |
| |
| self.ks = ks |
| self.dwconv = nn.Conv3d(dim, dim, kernel_size=self.ks, padding=self.ks//2, groups=dim) |
| self.norm = nn.BatchNorm3d(dim) |
| self.act = nn.GELU() |
| self.deploy = deploy |
|
|
| |
| if self.deploy == False: |
| alpha = a |
| BFM = np.zeros((dim, 1, self.ks, self.ks, self.ks)) |
| for i in range(self.ks): |
| for j in range(self.ks): |
| for k in range(self.ks): |
| point_1 = (i, j, k) |
| point_2 = (self.ks//2, self.ks//2, self.ks//2) |
| dist = distance.euclidean(point_1, point_2) |
| BFM[:, :, i, j, k] = alpha / (dist + alpha) |
| self.BFM = torch.from_numpy(BFM).type(torch.cuda.FloatTensor) |
|
|
| def forward(self, x): |
| |
| if self.deploy == False: |
| w_0 = self.dwconv.weight |
| w_1 = w_0 * self.BFM |
| self.dwconv.weight = torch.nn.Parameter(w_1) |
|
|
| |
| feat = self.dwconv(x) |
| feat = self.norm(feat) |
| feat = self.act(feat) |
|
|
| return feat |
|
|
|
|
| class repuxnet_conv(nn.Module): |
| """ |
| Args: |
| in_chans (int): Number of input image channels. Default: 3 |
| num_classes (int): Number of classes for classification head. Default: 1000 |
| depths (tuple(int)): Number of blocks at each stage. Default: [3, 3, 9, 3] |
| dims (int): Feature dimension at each stage. Default: [96, 192, 384, 768] |
| drop_path_rate (float): Stochastic depth rate. Default: 0. |
| layer_scale_init_value (float): Init value for Layer Scale. Default: 1e-6. |
| head_init_scale (float): Init scaling value for classifier weights and biases. Default: 1. |
| """ |
| def __init__(self, in_chans=1, depths=[2, 2, 2, 2], dims=[48, 96, 192, 384], ks=21, a=1, |
| drop_path_rate=0., layer_scale_init_value=1e-6, out_indices=[0, 1, 2, 3], deploy=False): |
| super().__init__() |
|
|
| self.downsample_layers = nn.ModuleList() |
| |
| |
| |
| |
| stem = nn.Sequential( |
| nn.Conv3d(in_chans, dims[0], kernel_size=7, stride=2, padding=3), |
| LayerNorm(dims[0], eps=1e-6, data_format="channels_first") |
| ) |
| self.downsample_layers.append(stem) |
| for i in range(3): |
| downsample_layer = nn.Sequential( |
| LayerNorm(dims[i], eps=1e-6, data_format="channels_first"), |
| nn.Conv3d(dims[i], dims[i+1], kernel_size=2, stride=2), |
| ) |
| self.downsample_layers.append(downsample_layer) |
|
|
| self.stages = nn.ModuleList() |
| dp_rates = [x.item() for x in torch.linspace(0, drop_path_rate, sum(depths))] |
| cur = 0 |
| self.deploy = deploy |
| for i in range(4): |
| stage = nn.Sequential( |
| *[repux_block(dim=dims[i], ks=ks, a=a, drop_path=dp_rates[cur + j], |
| layer_scale_init_value=layer_scale_init_value, deploy=self.deploy) for j in range(depths[i])] |
| ) |
| self.stages.append(stage) |
| cur += depths[i] |
|
|
| self.out_indices = out_indices |
|
|
| norm_layer = partial(LayerNorm, eps=1e-6, data_format="channels_first") |
| for i_layer in range(4): |
| layer = norm_layer(dims[i_layer]) |
| layer_name = f'norm{i_layer}' |
| self.add_module(layer_name, layer) |
|
|
| |
|
|
| def forward_features(self, x): |
| outs = [] |
| for i in range(4): |
| |
| |
| x = self.downsample_layers[i](x) |
| |
| x = self.stages[i](x) |
| |
| if i in self.out_indices: |
| norm_layer = getattr(self, f'norm{i}') |
| x_out = norm_layer(x) |
| outs.append(x_out) |
|
|
| return tuple(outs) |
|
|
| def forward(self, x): |
| x = self.forward_features(x) |
| return x |
|
|
|
|
