PathDualSwinUnet.py

import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.utils.checkpoint as checkpoint
import numpy as np
from timm.models.layers import DropPath, to_2tuple, trunc_normal_

from utils.checkpoint import load_checkpoint

from .arch_util.module import Attention, PreNorm, FeedForward
from monai.networks.blocks import UnetrUpBlock
import torch._dynamo

torch._dynamo.config.suppress_errors = True
groups = 32


class Mlp(nn.Module):
    """ Multilayer perceptron."""

    def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.):
        super().__init__()
        out_features = out_features or in_features
        hidden_features = hidden_features or in_features
        self.fc1 = nn.Linear(in_features, hidden_features).cuda()
        self.act = act_layer()
        self.fc2 = nn.Linear(hidden_features, out_features).cuda()
        self.drop = nn.Dropout(drop)

    def forward(self, x):
        x = self.fc1(x)
        x = self.act(x)
        x = self.drop(x)
        x = self.fc2(x)
        x = self.drop(x)
        return x


def window_partition(x, window_size):
    """
    Args:
        x: (B, H, W, C)
        window_size (int): window size
    Returns:
        windows: (num_windows*B, window_size, window_size, C)
    """
    B, H, W, C = x.shape
    x = x.view(B, H // window_size, window_size, W // window_size, window_size, C)
    windows = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(-1, window_size, window_size, C)
    return windows


def window_reverse(windows, window_size, H, W):
    """
    Args:
        windows: (num_windows*B, window_size, window_size, C)
        window_size (int): Window size
        H (int): Height of image
        W (int): Width of image
    Returns:
        x: (B, H, W, C)
    """
    B = int(windows.shape[0] / (H * W / window_size / window_size))
    x = windows.view(B, H // window_size, W // window_size, window_size, window_size, -1)
    x = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(B, H, W, -1)
    return x


class WindowAttention(nn.Module):
    """ Window based multi-head self attention (W-MSA) module with relative position bias.
    It supports both of shifted and non-shifted window.
    Args:
        dim (int): Number of input channels.
        window_size (tuple[int]): The height and width of the window.
        num_heads (int): Number of attention heads.
        qkv_bias (bool, optional):  If True, add a learnable bias to query, key, value. Default: True
        qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set
        attn_drop (float, optional): Dropout ratio of attention weight. Default: 0.0
        proj_drop (float, optional): Dropout ratio of output. Default: 0.0
    """

    def __init__(self, dim, window_size, num_heads, qkv_bias=True, qk_scale=None, attn_drop=0., proj_drop=0.):

        super().__init__()
        self.dim = dim
        self.window_size = window_size  # Wh, Ww
        self.num_heads = num_heads
        head_dim = dim // num_heads
        self.scale = qk_scale or head_dim ** -0.5

        # define a parameter table of relative position bias
        self.relative_position_bias_table = nn.Parameter(
            torch.zeros((2 * window_size[0] - 1) * (2 * window_size[1] - 1), num_heads))  # 2*Wh-1 * 2*Ww-1, nH

        # get pair-wise relative position index for each token inside the window
        coords_h = torch.arange(self.window_size[0])
        coords_w = torch.arange(self.window_size[1])
        coords = torch.stack(torch.meshgrid([coords_h, coords_w]))  # 2, Wh, Ww
        coords_flatten = torch.flatten(coords, 1)  # 2, Wh*Ww
        relative_coords = coords_flatten[:, :, None] - coords_flatten[:, None, :]  # 2, Wh*Ww, Wh*Ww
        relative_coords = relative_coords.permute(1, 2, 0).contiguous()  # Wh*Ww, Wh*Ww, 2
        relative_coords[:, :, 0] += self.window_size[0] - 1  # shift to start from 0
        relative_coords[:, :, 1] += self.window_size[1] - 1
        relative_coords[:, :, 0] *= 2 * self.window_size[1] - 1
        relative_position_index = relative_coords.sum(-1)  # Wh*Ww, Wh*Ww
        self.register_buffer("relative_position_index", relative_position_index)

        self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)
        self.attn_drop = nn.Dropout(attn_drop)
        self.proj = nn.Linear(dim, dim)
        self.proj_drop = nn.Dropout(proj_drop)

        trunc_normal_(self.relative_position_bias_table, std=.02)
        self.softmax = nn.Softmax(dim=-1)

    def forward(self, x, mask=None):
        """ Forward function.
        Args:
            x: input features with shape of (num_windows*B, N, C)
            mask: (0/-inf) mask with shape of (num_windows, Wh*Ww, Wh*Ww) or None
        """
        B_, N, C = x.shape
        qkv = self.qkv(x).reshape(B_, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
        q, k, v = qkv[0], qkv[1], qkv[2]  # make torchscript happy (cannot use tensor as tuple)

        q = q * self.scale
        attn = (q @ k.transpose(-2, -1))

        relative_position_bias = self.relative_position_bias_table[self.relative_position_index.view(-1)].view(
            self.window_size[0] * self.window_size[1], self.window_size[0] * self.window_size[1], -1)  # Wh*Ww,Wh*Ww,nH
        relative_position_bias = relative_position_bias.permute(2, 0, 1).contiguous()  # nH, Wh*Ww, Wh*Ww
        attn = attn + relative_position_bias.unsqueeze(0)

        if mask is not None:
            nW = mask.shape[0]
            attn = attn.view(B_ // nW, nW, self.num_heads, N, N) + mask.unsqueeze(1).unsqueeze(0)
            attn = attn.view(-1, self.num_heads, N, N)
            attn = self.softmax(attn)
        else:
            attn = self.softmax(attn)

        attn = self.attn_drop(attn)

        x = (attn @ v).transpose(1, 2).reshape(B_, N, C)
        x = self.proj(x)
        x = self.proj_drop(x)
        return x


class SwinTransformerBlock(nn.Module):
    """ Swin Transformer Block.
    Args:
        dim (int): Number of input channels.
        num_heads (int): Number of attention heads.
        window_size (int): Window size.
        shift_size (int): Shift size for SW-MSA.
        mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.
        qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True
        qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set.
        drop (float, optional): Dropout rate. Default: 0.0
        attn_drop (float, optional): Attention dropout rate. Default: 0.0
        drop_path (float, optional): Stochastic depth rate. Default: 0.0
        act_layer (nn.Module, optional): Activation layer. Default: nn.GELU
        norm_layer (nn.Module, optional): Normalization layer.  Default: nn.LayerNorm
    """

    def __init__(self, dim, num_heads, window_size=7, shift_size=0,
                 mlp_ratio=4., qkv_bias=True, qk_scale=None, drop=0., attn_drop=0., drop_path=0.,
                 act_layer=nn.GELU, norm_layer=nn.LayerNorm):
        super().__init__()
        self.dim = dim
        self.num_heads = num_heads
        self.window_size = window_size
        self.shift_size = shift_size
        self.mlp_ratio = mlp_ratio

        assert 0 <= self.shift_size < self.window_size, "shift_size must in 0-window_size"

        self.norm1 = norm_layer(dim)
        self.attn = WindowAttention(
            dim, window_size=to_2tuple(self.window_size), num_heads=num_heads,
            qkv_bias=qkv_bias, qk_scale=qk_scale, attn_drop=attn_drop, proj_drop=drop)

        self.drop_path = DropPath(drop_path) if drop_path > 0. else nn.Identity()
        self.norm2 = norm_layer(dim)
        mlp_hidden_dim = int(dim * mlp_ratio)
        self.mlp = Mlp(in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop)

        self.H = None
        self.W = None

    def forward(self, x, mask_matrix):
        """ Forward function.
        Args:
            x: Input feature, tensor size (B, H*W, C).
            x_size: Spatial resolution of the input feature.
            mask_matrix: Attention mask for cyclic shift, calculated in basic layer
        """
        B, L, C = x.shape
        H, W = self.H, self.W
        assert L == H * W, "input feature has wrong size"

        shortcut = x
        x = self.norm1(x)
        x = x.view(B, H, W, C)

        # pad feature maps to multiples of window size
        pad_l = pad_t = 0
        pad_r = (self.window_size - W % self.window_size) % self.window_size
        pad_b = (self.window_size - H % self.window_size) % self.window_size
        x = F.pad(x, (0, 0, pad_l, pad_r, pad_t, pad_b))
        _, Hp, Wp, _ = x.shape

        # cyclic shift
        if self.shift_size > 0:
            shifted_x = torch.roll(x, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2))
            attn_mask = mask_matrix
        else:
            shifted_x = x
            attn_mask = None

        # partition windows
        x_windows = window_partition(shifted_x, self.window_size)  # nW*B, window_size, window_size, C
        x_windows = x_windows.view(-1, self.window_size * self.window_size, C)  # nW*B, window_size*window_size, C

        # W-MSA/SW-MSA
        attn_windows = self.attn(x_windows, mask=attn_mask)  # nW*B, window_size*window_size, C

        # merge windows
        attn_windows = attn_windows.view(-1, self.window_size, self.window_size, C)

        # reverse cyclic shift
        shifted_x = window_reverse(attn_windows, self.window_size, Hp, Wp)  # B H' W' C
        if self.shift_size > 0:
            x = torch.roll(shifted_x, shifts=(self.shift_size, self.shift_size), dims=(1, 2))
        else:
            x = shifted_x

        if pad_r > 0 or pad_b > 0:
            x = x[:, :H, :W, :].contiguous()

        x = x.view(B, H * W, C)

        # FFN
        x = shortcut + self.drop_path(x)
        x = x + self.drop_path(self.mlp(self.norm2(x)))

        return x


class PatchMerging(nn.Module):
    """ Patch Merging Layer
    Args:
        input_resolution (tuple[int]): Resolution of input feature.
        dim (int): Number of input channels.
        norm_layer (nn.Module, optional): Normalization layer.  Default: nn.LayerNorm
    """

    def __init__(self, dim, norm_layer=nn.LayerNorm):
        super().__init__()
        self.dim = dim
        self.reduction = nn.Linear(4 * dim, 2 * dim, bias=False)
        self.norm = norm_layer(4 * dim)

    def forward(self, x, H, W):
        """ Forward function.
        Args:
            x: Input feature, tensor size (B, H*W, C).
            H, W: Spatial resolution of the input feature.
        """
        B, L, C = x.shape
        assert L == H * W, "input feature has wrong size"

        x = x.view(B, H, W, C)

        # padding
        pad_input = (H % 2 == 1) or (W % 2 == 1)
        if pad_input:
            x = F.pad(x, (0, 0, 0, W % 2, 0, H % 2))

        x0 = x[:, 0::2, 0::2, :]  # B H/2 W/2 C
        x1 = x[:, 1::2, 0::2, :]  # B H/2 W/2 C
        x2 = x[:, 0::2, 1::2, :]  # B H/2 W/2 C
        x3 = x[:, 1::2, 1::2, :]  # B H/2 W/2 C
        x = torch.cat([x0, x1, x2, x3], -1)  # B H/2 W/2 4*C
        x = x.view(B, -1, 4 * C)  # B H/2*W/2 4*C

        x = self.norm(x)
        x = self.reduction(x)

        return x


class BasicLayer(nn.Module):
    """ A basic Swin Transformer layer for one stage.
    Args:
        dim (int): Number of feature channels
        depth (int): Depths of this stage.
        num_heads (int): Number of attention head.
        window_size (int): Local window size. Default: 7.
        mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. Default: 4.
        qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True
        qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set.
        drop (float, optional): Dropout rate. Default: 0.0
        attn_drop (float, optional): Attention dropout rate. Default: 0.0
        drop_path (float | tuple[float], optional): Stochastic depth rate. Default: 0.0
        norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm
        downsample (nn.Module | None, optional): Downsample layer at the end of the layer. Default: None
        use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False.
    """

    def __init__(self, dim, depth, num_heads, window_size=7,
                 mlp_ratio=4., qkv_bias=True, qk_scale=None, drop=0., attn_drop=0.,
                 drop_path=0., norm_layer=nn.LayerNorm, downsample=None, use_checkpoint=False,
                 up=True):
        super().__init__()
        self.dim = dim
        self.window_size = window_size
        self.shift_size = window_size // 2
        self.depth = depth
        self.use_checkpoint = use_checkpoint
        self.up = up

        # build blocks
        self.blocks = nn.ModuleList([
            SwinTransformerBlock(
                dim=dim,
                num_heads=num_heads,
                window_size=window_size,
                shift_size=0 if (i % 2 == 0) else window_size // 2,
                mlp_ratio=mlp_ratio,
                qkv_bias=qkv_bias,
                qk_scale=qk_scale,
                drop=drop,
                attn_drop=attn_drop,
                drop_path=drop_path[i] if isinstance(drop_path, list) else drop_path,
                norm_layer=norm_layer)
            for i in range(depth)])

        # patch merging layer
        if downsample is not None:
            self.downsample = downsample(dim=dim, norm_layer=norm_layer)
        else:
            self.downsample = None

    def forward(self, x, H, W):
        """ Forward function.
        Args:
            x: Input feature, tensor size (B, H*W, C).
            x_size: Spatial resolution of the input feature.
        """
        # calculate attention mask for SW-MSA
        Hp = int(np.ceil(H / self.window_size)) * self.window_size
        Wp = int(np.ceil(W / self.window_size)) * self.window_size
        img_mask = torch.zeros((1, Hp, Wp, 1), device=x.device)  # 1 Hp Wp 1
        h_slices = (slice(0, -self.window_size),
                    slice(-self.window_size, -self.shift_size),
                    slice(-self.shift_size, None))
        w_slices = (slice(0, -self.window_size),
                    slice(-self.window_size, -self.shift_size),
                    slice(-self.shift_size, None))
        cnt = 0
        for h in h_slices:
            for w in w_slices:
                img_mask[:, h, w, :] = cnt
                cnt += 1

        mask_windows = window_partition(img_mask, self.window_size)  # nW, window_size, window_size, 1
        mask_windows = mask_windows.view(-1, self.window_size * self.window_size)
        attn_mask = mask_windows.unsqueeze(1) - mask_windows.unsqueeze(2)
        attn_mask = attn_mask.masked_fill(attn_mask != 0, float(-100.0)).masked_fill(attn_mask == 0, float(0.0))

        for blk in self.blocks:
            blk.H, blk.W = H, W
            if self.use_checkpoint:
                x = checkpoint.checkpoint(blk, x, attn_mask)
            else:
                x = blk(x, attn_mask)
        if self.downsample is not None:
            x_down = self.downsample(x, H, W)
            if self.up:
                Wh, Ww = (H + 1) // 2, (W + 1) // 2
            else:
                Wh, Ww = H * 2, W * 2
            return x, H, W, x_down, Wh, Ww
        else:
            return x, H, W, x, H, W


class PatchEmbed(nn.Module):

    def __init__(self, patch_size=4, in_chans=3, embed_dim=96, norm_layer=None):
        super().__init__()
        patch_size = to_2tuple(patch_size)
        self.patch_size = patch_size

        self.in_chans = in_chans
        self.embed_dim = embed_dim

        self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size, stride=patch_size)
        if norm_layer is not None:
            self.norm = norm_layer(embed_dim)
        else:
            self.norm = None

    def forward(self, x):
        """Forward function."""
        # padding
        B, C, H, W = x.size()
        if W % self.patch_size[1] != 0:
            x = F.pad(x, (0, self.patch_size[1] - W % self.patch_size[1]))
        if H % self.patch_size[0] != 0:
            x = F.pad(x, (0, 0, 0, self.patch_size[0] - H % self.patch_size[0]))
        x = self.proj(x)  # B C Wh Ww
        if self.norm is not None:
            Wh, Ww = x.size(2), x.size(3)
            x = x.flatten(2).transpose(1, 2)
            x = self.norm(x)
            x = x.transpose(1, 2).view(-1, self.embed_dim, Wh, Ww)

        return x


class PatchUnEmbed(nn.Module):
    r""" Image to Patch Unembedding

    Args:
        patch_size (int): Patch token size. Default: 4.
        in_chans (int): Number of input image channels. Default: 3.
        embed_dim (int): Number of linear projection output channels. Default: 96.
    """

    def __init__(self, patch_size=4, in_chans=3, embed_dim=96):
        super().__init__()
        patch_size = to_2tuple(patch_size)
        self.patch_size = patch_size

        self.in_chans = in_chans
        self.embed_dim = embed_dim

        self.proj = nn.ConvTranspose2d(embed_dim, embed_dim, kernel_size=patch_size, stride=patch_size)

    def forward(self, x, H, W):
        B, C, L = x.shape
        x = self.upscale(x)
        x = x.transpose(1, 2)
        x = self.proj(x)
        x = x.view(B, self.embed_dim, H, W)  # B Ph*Pw C
        return x

    def flops(self):
        flops = 0
        return flops


class SwinTransformer(nn.Module):
    """ Swin Transformer backbone.
        A PyTorch impl of : `Swin Transformer: Hierarchical Vision Transformer using Shifted Windows`  -
          https://arxiv.org/pdf/2103.14030
    Args:
        pretrain_img_size (int): Input image size for training the pretrained model,
            used in absolute postion embedding. Default 224.
        patch_size (int | tuple(int)): Patch size. Default: 4.
        in_chans (int): Number of input image channels. Default: 3.
        embed_dim (int): Number of linear projection output channels. Default: 96.
        depths (tuple[int]): Depths of each Swin Transformer stage.
        num_heads (tuple[int]): Number of attention head of each stage.
        window_size (int): Window size. Default: 7.
        mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. Default: 4.
        qkv_bias (bool): If True, add a learnable bias to query, key, value. Default: True
        qk_scale (float): Override default qk scale of head_dim ** -0.5 if set.
        drop_rate (float): Dropout rate.
        attn_drop_rate (float): Attention dropout rate. Default: 0.
        drop_path_rate (float): Stochastic depth rate. Default: 0.2.
        norm_layer (nn.Module): Normalization layer. Default: nn.LayerNorm.
        ape (bool): If True, add absolute position embedding to the patch embedding. Default: False.
        patch_norm (bool): If True, add normalization after patch embedding. Default: True.
        out_indices (Sequence[int]): Output from which stages.
        frozen_stages (int): Stages to be frozen (stop grad and set eval mode).
            -1 means not freezing any parameters.
        use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False.
    """

    def __init__(self, pretrain_img_size=224, patch_size=4, in_chans=3, embed_dim=128,
                 depths=[2, 2, 6, 2], num_heads=[3, 6, 12, 24], window_size=7, mlp_ratio=4., qkv_bias=True,
                 qk_scale=None,
                 drop_rate=0., attn_drop_rate=0., drop_path_rate=0.1, norm_layer=nn.LayerNorm, ape=False,
                 patch_norm=True, out_indices=(0, 1, 2, 3), frozen_stages=-1, use_checkpoint=False):
        super().__init__()

        self.pretrain_img_size = pretrain_img_size
        self.num_layers = len(depths)
        self.embed_dim = embed_dim
        self.ape = ape
        self.mlp_ratio = mlp_ratio
        self.patch_norm = patch_norm
        self.out_indices = out_indices
        self.frozen_stages = frozen_stages

        # shallow feature extraction at the beginning
        self.conv_first = nn.Conv2d(in_chans, embed_dim, 3, 1, 1)

        # split image into non-overlapping patches
        self.patch_embed = PatchEmbed(
            patch_size=patch_size, in_chans=embed_dim, embed_dim=embed_dim,
            norm_layer=norm_layer if self.patch_norm else None)

        # merge non-overlapping patches into image
        self.patch_unembed = PatchUnEmbed(patch_size=patch_size, in_chans=embed_dim, embed_dim=embed_dim)

        # absolute position embedding
        if self.ape:
            pretrain_img_size = to_2tuple(pretrain_img_size)
            patch_size = to_2tuple(patch_size)
            patches_resolution = [pretrain_img_size[0] // patch_size[0], pretrain_img_size[1] // patch_size[1]]

            self.absolute_pos_embed = nn.Parameter(
                torch.zeros(1, embed_dim, patches_resolution[0], patches_resolution[1]))
            trunc_normal_(self.absolute_pos_embed, std=.02)

        self.pos_drop = nn.Dropout(p=drop_rate)

        # stochastic depth
        dpr = [x.item() for x in torch.linspace(0, drop_path_rate, sum(depths))]  # stochastic depth decay rule

        # build layers
        self.layers = nn.ModuleList()
        for i_layer in range(self.num_layers):
            layer = BasicLayer(
                dim=int(embed_dim * 2 ** i_layer),
                depth=depths[i_layer],
                num_heads=num_heads[i_layer],
                window_size=window_size,
                mlp_ratio=mlp_ratio,
                qkv_bias=qkv_bias,
                qk_scale=qk_scale,
                drop=drop_rate,
                attn_drop=attn_drop_rate,
                drop_path=dpr[sum(depths[:i_layer]):sum(depths[:i_layer + 1])],
                norm_layer=norm_layer,
                downsample=PatchMerging if (i_layer < self.num_layers - 1) else None,
                use_checkpoint=use_checkpoint)
            self.layers.append(layer)

        num_features = [int(embed_dim * 2 ** i) for i in range(self.num_layers)]
        self.num_features = num_features

        # add a norm layer for each output
        for i_layer in out_indices:
            layer = norm_layer(num_features[i_layer])
            layer_name = f'norm{i_layer}'
            self.add_module(layer_name, layer)
        self.avgpool = nn.AdaptiveAvgPool1d(1)
        self.head = nn.Linear(self.num_features[3], 1)
        self._freeze_stages()

    def _freeze_stages(self):
        if self.frozen_stages >= 0:
            self.patch_embed.eval()
            for param in self.patch_embed.parameters():
                param.requires_grad = False

        if self.frozen_stages >= 1 and self.ape:
            self.absolute_pos_embed.requires_grad = False

        if self.frozen_stages >= 2:
            self.pos_drop.eval()
            for i in range(0, self.frozen_stages - 1):
                m = self.layers[i]
                m.eval()
                for param in m.parameters():
                    param.requires_grad = False

    def init_weights(self, pretrained=None):
        """Initialize the weights in backbone.
        Args:
            pretrained (str, optional): Path to pre-trained weights.
                Defaults to None.
        """

        def _init_weights(m):
            if isinstance(m, nn.Linear):
                trunc_normal_(m.weight, std=.02)
                if isinstance(m, nn.Linear) and m.bias is not None:
                    nn.init.constant_(m.bias, 0)
            elif isinstance(m, nn.LayerNorm):
                nn.init.constant_(m.bias, 0)
                nn.init.constant_(m.weight, 1.0)

        if isinstance(pretrained, str):
            self.apply(_init_weights)
            load_checkpoint(self, pretrained, strict=False)
        elif pretrained is None:
            self.apply(_init_weights)
        else:
            raise TypeError('pretrained must be a str or None')

    def forward(self, x):
        """Forward function."""
        x = self.conv_first(x)
        x = self.patch_embed(x)
        Wh_orig, Ww_orig = x.shape[2], x.shape[3]
        Wh = Wh_orig
        Ww = Ww_orig

        if self.ape:
            # interpolate the position embedding to the corresponding size
            absolute_pos_embed = F.interpolate(self.absolute_pos_embed, size=(Wh, Ww), mode='bicubic')
            x = (x + absolute_pos_embed).flatten(2).transpose(1, 2)  # B Wh*Ww C
        else:
            x = x.flatten(2).transpose(1, 2)
        x = self.pos_drop(x)

        outs = []
        for i in range(self.num_layers):
            layer = self.layers[i]
            x_out, H, W, x, Wh, Ww = layer(x, Wh, Ww)

            if i in self.out_indices:
                norm_layer = getattr(self, f'norm{i}')
                x_out = norm_layer(x_out)

                out = x_out.view(-1, H, W, self.num_features[i]).permute(0, 3, 1, 2).contiguous()
                outs.append(out)

        outs_binary = self.avgpool(outs[3].flatten(2))
        outs_binary = torch.flatten(outs_binary, 1)
        outs_binary = self.head(outs_binary)
        return outs, outs_binary

    def train(self, mode=True):
        """Convert the model into training mode while keep layers freezed."""
        super(SwinTransformer, self).train(mode)
        self._freeze_stages()


class conv_block(nn.Module):
    """
    Convolution Block
    """

    def __init__(self, in_ch, out_ch):
        super(conv_block, self).__init__()
        self.conv = nn.Sequential(
            nn.MaxPool2d(kernel_size=2, stride=2),
            nn.Conv2d(in_ch, out_ch, kernel_size=3, stride=1, padding=1, bias=True),
            nn.GroupNorm(num_channels=out_ch, num_groups=groups),
            nn.ReLU(inplace=True),
            nn.Conv2d(out_ch, out_ch, kernel_size=3, stride=1, padding=1, bias=True),
            nn.GroupNorm(num_channels=out_ch, num_groups=groups),
            nn.ReLU(inplace=True))

    def forward(self, x):
        x = self.conv(x)
        return x


class Conv_block(nn.Module):
    """
    Convolution Block
    """

    def __init__(self, in_ch, out_ch):
        super(Conv_block, self).__init__()
        self.conv = nn.Sequential(
            nn.Conv2d(in_ch, out_ch, kernel_size=3, stride=1, padding=1, bias=True),
            nn.GroupNorm(num_channels=out_ch, num_groups=groups),
            nn.ReLU(inplace=True),
            nn.Conv2d(out_ch, out_ch, kernel_size=3, stride=1, padding=1, bias=True),
            nn.GroupNorm(num_channels=out_ch, num_groups=groups),
            nn.ReLU(inplace=True))

    def forward(self, x):
        x = self.conv(x)
        return x


class Transformer(nn.Module):
    def __init__(self, dim, depth, heads, dim_head, mlp_dim, dropout=0.):
        super().__init__()
        self.layers = nn.ModuleList([])
        for _ in range(depth):
            self.layers.append(nn.ModuleList([
                PreNorm(dim, Attention(dim, heads=heads, dim_head=dim_head, dropout=dropout)),
                PreNorm(dim, FeedForward(dim, mlp_dim, dropout=dropout))
            ]))

    def forward(self, x):
        for attn, ff in self.layers:
            x = attn(x) + x
            x = ff(x) + x
        return x


class Cross_Att(nn.Module):
    def __init__(self, dim_s, dim_l):
        super().__init__()
        self.transformer_s = Transformer(dim=dim_s, depth=1, heads=3, dim_head=32, mlp_dim=128)
        self.transformer_l = Transformer(dim=dim_l, depth=1, heads=1, dim_head=64, mlp_dim=256)
        self.norm_s = nn.LayerNorm(dim_s)
        self.norm_l = nn.LayerNorm(dim_l)
        self.avgpool = nn.AdaptiveAvgPool1d(1)
        self.linear_s = nn.Linear(dim_s, dim_l)
        self.linear_l = nn.Linear(dim_l, dim_s)

    def forward(self, e, r):
        b_e, c_e, h_e, w_e = e.shape
        e = e.reshape(b_e, c_e, -1).permute(0, 2, 1)
        b_r, c_r, h_r, w_r = r.shape
        r = r.reshape(b_r, c_r, -1).permute(0, 2, 1)
        e_t = torch.flatten(self.avgpool(self.norm_l(e).transpose(1, 2)), 1)
        r_t = torch.flatten(self.avgpool(self.norm_s(r).transpose(1, 2)), 1)
        e_t = self.linear_l(e_t).unsqueeze(1)
        r_t = self.linear_s(r_t).unsqueeze(1)
        r = self.transformer_s(torch.cat([e_t, r], dim=1))[:, 1:, :]
        e = self.transformer_l(torch.cat([r_t, e], dim=1))[:, 1:, :]
        e = e.permute(0, 2, 1).reshape(b_e, c_e, h_e, w_e)
        r = r.permute(0, 2, 1).reshape(b_r, c_r, h_r, w_r)
        return e, r


class DsTransUNet(nn.Module):
    def __init__(self, dim, **kwargs):
        super().__init__()
        in_channels = kwargs['in_channels']
        out_channels = kwargs['out_channels']
        self.encoder = SwinTransformer(depths=[2, 2, 18, 2], num_heads=[4, 8, 16, 32], drop_path_rate=0.5, patch_size=4,
                                       embed_dim=128)
        self.encoder2 = SwinTransformer(depths=[2, 2, 6, 2], num_heads=[3, 6, 12, 24], drop_path_rate=0.2, patch_size=8,
                                        embed_dim=96)
        if kwargs['pretrained']:
            self.encoder.init_weights(os.path.join(kwargs["pretrained_model_path"],
                                                   'swin_base_mim.pth'))
            self.encoder2.init_weights(os.path.join(kwargs["pretrained_model_path"],
                                                    'swin_tiny_mim.pth'))

        self.decoder5 = UnetrUpBlock(spatial_dims=2, in_channels=8 * dim, out_channels=4 * dim, kernel_size=3,
                                     upsample_kernel_size=2, norm_name="instance", res_block=True, )
        self.decoder4 = UnetrUpBlock(spatial_dims=2, in_channels=4 * dim, out_channels=2 * dim, kernel_size=3,
                                     upsample_kernel_size=2, norm_name="instance", res_block=True, )
        self.decoder3 = UnetrUpBlock(spatial_dims=2, in_channels=2 * dim, out_channels=dim, kernel_size=3,
                                     upsample_kernel_size=2, norm_name="instance", res_block=True, )
        self.decoder2 = UnetrUpBlock(spatial_dims=2, in_channels=dim, out_channels=dim // 2, kernel_size=3,
                                     upsample_kernel_size=2, norm_name="instance", res_block=True, )
        self.decoder1 = UnetrUpBlock(spatial_dims=2, in_channels=dim // 2, out_channels=dim // 4, kernel_size=3,
                                     upsample_kernel_size=2, norm_name="instance", res_block=True, )

        self.down1 = nn.Conv2d(in_channels, dim // 4, kernel_size=1, stride=1, padding=0)
        self.down2 = conv_block(dim // 4, dim // 2)
        self.final = nn.Conv2d(dim // 4, out_channels, kernel_size=1, stride=1, padding=0)

        self.loss1 = nn.Sequential(
            nn.Conv2d(dim * 8, out_channels, kernel_size=1, stride=1, padding=0),
            nn.ReLU(),
            nn.Upsample(scale_factor=32)
        )

        self.loss2 = nn.Sequential(
            nn.Conv2d(dim, out_channels, kernel_size=1, stride=1, padding=0),
            nn.ReLU(),
            nn.Upsample(scale_factor=4)
        )
        dim_s = 96
        dim_l = 128
        self.m1 = nn.Upsample(scale_factor=2)
        self.m2 = nn.Upsample(scale_factor=4)
        tb = dim_s + dim_l
        self.change1 = Conv_block(tb, dim)
        self.change2 = Conv_block(tb * 2, dim * 2)
        self.change3 = Conv_block(tb * 4, dim * 4)
        self.change4 = Conv_block(tb * 8, dim * 8)
        self.cross_att_1 = Cross_Att(dim_s * 1, dim_l * 1)
        self.cross_att_2 = Cross_Att(dim_s * 2, dim_l * 2)
        self.cross_att_3 = Cross_Att(dim_s * 4, dim_l * 4)
        self.cross_att_4 = Cross_Att(dim_s * 8, dim_l * 8)

    def forward(self, x):
        out, loss3 = self.encoder(x)
        out2, loss4 = self.encoder2(x)
        e11, e12, e13, e14 = out[0], out[1], out[2], out[3]
        e21, e22, e23, e24 = out2[0], out2[1], out2[2], out2[3]
        e11, e21 = self.cross_att_1(e11, e21)
        e12, e22 = self.cross_att_2(e12, e22)
        e13, e23 = self.cross_att_3(e13, e23)
        e14, e24 = self.cross_att_4(e14, e24)
        e1 = torch.cat([e11, self.m1(e21)], 1)
        e2 = torch.cat([e12, self.m1(e22)], 1)
        e3 = torch.cat([e13, self.m1(e23)], 1)
        e4 = torch.cat([e14, self.m1(e24)], 1)
        e1 = self.change1(e1)
        e2 = self.change2(e2)
        e3 = self.change3(e3)
        e4 = self.change4(e4)
        loss1 = self.loss1(e4)
        ds1 = self.down1(x)
        ds2 = self.down2(ds1)
        d1 = self.decoder5(e4, e3)
        d2 = self.decoder4(d1, e2)
        d3 = self.decoder3(d2, e1)
        loss2 = self.loss2(d3)

        d4 = self.decoder2(d3, ds2)
        d5 = self.decoder1(d4, ds1)
        o = self.final(d5)
        if not self.training:
            loss3 = loss3.unsqueeze(1).repeat(1, 1, o.shape[2], o.shape[3]).movedim(0, 1)
            loss4 = loss4.unsqueeze(1).repeat(1, 1, o.shape[2], o.shape[3]).movedim(0, 1)

        return o, loss1, loss2, (loss3 + loss4) / 2


if __name__ == '__main__':
    print('#### Test Case ###')
    from torch.autograd import Variable

    keywords = {'in_channels': 3, 'out_channels': 1, 'pretrained': False}

    x = Variable(torch.rand(2, 3, 256, 256)).cuda()
    model = DsTransUNet(128, in_channels=3, out_channels=1, pretrained=False).cuda()
    print("Input shape:", x.shape)
    y = model(x)
    print('Output shape:', y[-1].shape)