lib/backbone_ppm.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 .mmcv_custom import load_checkpoint
from mmseg.utils import get_root_logger


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)
        self.act = act_layer()
        self.fc2 = nn.Linear(hidden_features, out_features)
        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)  # cat op
        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).
            H, W: Spatial resolution of the input feature.
            mask_matrix: Attention mask for cyclic shift.
        """
        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)
        shifted_x = window_reverse(attn_windows, self.window_size, Hp, Wp)  # B H' W' C

        # reverse cyclic shift
        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 feed-forward network
        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:
        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 PatchEmbed(nn.Module):
    """ Image to Patch Embedding

    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.
        norm_layer (nn.Module, optional): Normalization layer. Default: None
    """

    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
        _, _, 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 MultiModalSwinTransformer(nn.Module):
    def __init__(self,
                 pretrain_img_size=224,
                 patch_size=4,
                 in_chans=3,
                 embed_dim=96,
                 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=0.2,
                 norm_layer=nn.LayerNorm,
                 ape=False,
                 patch_norm=True,
                 out_indices=(0, 1, 2, 3),
                 frozen_stages=-1,
                 use_checkpoint=False,
                 num_heads_fusion=[1, 1, 1, 1],
                 fusion_drop=0.0
                 ):
        super().__init__()

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

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

        # 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 = MMBasicLayer(
                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,
                num_heads_fusion=num_heads_fusion[i_layer],
                fusion_drop=fusion_drop
            )
            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._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)
            logger = get_root_logger()
            load_checkpoint(self, pretrained, strict=('upernet' in pretrained), logger=logger)
        elif pretrained is None:
            self.apply(_init_weights)
        else:
            raise TypeError('pretrained must be a str or None')

    def forward(self, x, l, l_mask):
        """Forward function."""
        x = self.patch_embed(x)

        Wh, Ww = x.size(2), x.size(3)
        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, l, l_mask)

            if i in self.out_indices:
                norm_layer = getattr(self, f'norm{i}')
                x_out = norm_layer(x_out)  # output of a Block has shape (B, H*W, dim)

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

        return tuple(outs)

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

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_first"):
        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] * x + self.bias[:, None, None]
            return x


class MMBasicLayer(nn.Module):
    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,
                 num_heads_fusion=1,
                 fusion_drop=0.0
                 ):
        super().__init__()
        self.window_size = window_size
        self.shift_size = window_size // 2
        self.depth = depth
        self.use_checkpoint = use_checkpoint
        self.dim = dim

        # 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)])

        # fuse before downsampling
        self.fusion = PWAM(dim,  # both the visual input and for combining, num of channels
                           dim,  # v_in
                           768,  # l_in
                           dim,  # key
                           dim,  # value
                           num_heads=num_heads_fusion,
                           dropout=fusion_drop)

        self.res_gate = nn.Sequential(
            nn.Linear(dim, dim, bias=False),
            nn.GELU(),
            nn.Linear(dim, dim, bias=False),
            nn.Tanh()
        )

        self.psizes = [1,2,3,6]
        reduction_dim = dim // 4
        self.pyramids = nn.ModuleList()
        self.fusions = nn.ModuleList()
        self.mixer = nn.Sequential(
            nn.Linear(dim*2, dim),
            nn.LayerNorm(dim),
            nn.Linear(dim, dim),
            nn.GELU()
        )

        #self.res_gates = nn.ModuleList()
        for p in self.psizes:
            self.pyramids.append(
                #nn.Sequential(
                #    #nn.AdaptiveAvgPool2d(p),
                #    nn.Conv2d(dim, reduction_dim, kernel_size=p, padding=p//2, bias=False),
                #    nn.BatchNorm2d(reduction_dim),
                #    nn.ReLU(inplace=True)
                #)
                nn.Sequential(
                    nn.AdaptiveAvgPool2d(p),
                    nn.Conv2d(dim, dim*4, kernel_size=1, bias=False),
                    #nn.BatchNorm2d(reduction_dim),
                    LayerNorm(dim*4),
                    nn.Conv2d(dim*4, dim, kernel_size=1, bias=False),
                    nn.GELU(),
                    nn.Conv2d(dim, reduction_dim, kernel_size=1, bias=False),
                )
            )
            self.fusions.append(
                PWAM(reduction_dim,  # both the visual input and for combining, num of channels
                     reduction_dim,  # v_in
                     768,  # l_in
                     reduction_dim,  # key
                     reduction_dim,  # value
                     num_heads=num_heads_fusion,
                     dropout=fusion_drop)
            )
            self.reduction_dim = reduction_dim
            #self.res_gates.append(
            #    nn.Sequential(
            #        nn.Linear(reduction_dim, reduction_dim, bias=False),
            #        nn.ReLU(),
            #        nn.Linear(reduction_dim, reduction_dim, bias=False),
            #        nn.Tanh()
            #    )
            #)


        # patch merging layer
        if downsample is not None:
            self.downsample = downsample(dim=dim, norm_layer=norm_layer)
        else:
            self.downsample = None
        # initialize the gate to 0
        nn.init.zeros_(self.res_gate[0].weight)
        nn.init.zeros_(self.res_gate[2].weight)

    def forward(self, x, H, W, l, l_mask):
        """ Forward function.

        Args:
            x: Input feature, tensor size (B, H*W, C).
            H, W: 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)  # output of a Block has shape (B, H*W, dim)

        out = []

        #torch.Size([2, 32, 1, 1])
        #torch.Size([2, 1, 32])
        #P3WAM
        x_reshape = x.permute(0,2,1).view(x.shape[0], x.shape[2], H, W)
        x_size = x_reshape.size()
        for i, p in enumerate(self.psizes):
            px = self.pyramids[i](x_reshape)
            px = px.flatten(2).permute(0,2,1)
            #print(px.shape)
            px_residual = self.fusions[i](px, l, l_mask)
            px_residual = px_residual.permute(0,2,1).view(x.shape[0], self.reduction_dim , p, p)
            #print(px_residual.shape)
            out.append(F.interpolate(px_residual, x_size[2:], mode='bilinear', align_corners=True).flatten(2).permute(0,2,1))

        # PWAM fusion
        #x_residual = self.fusion(x, l, l_mask)
        ## apply a gate on the residual
        #x = x + (self.res_gate(x_residual) * x_residual)

        # PWAM fusion
        x_residual = self.fusion(x, l, l_mask)
        out.append(x_residual)
        # apply a gate on the residual
        x = x + (self.res_gate(x_residual) * x_residual)

        #print('---')
        #for o in out:
        #    print(o.shape)


        x_residual = self.mixer(torch.cat(out, dim =2))

        if self.downsample is not None:
            x_down = self.downsample(x, H, W)
            Wh, Ww = (H + 1) // 2, (W + 1) // 2
            return x_residual, H, W, x_down, Wh, Ww
        else:
            return x_residual, H, W, x, H, W


class PWAM(nn.Module):
    def __init__(self, dim, v_in_channels, l_in_channels, key_channels, value_channels, num_heads=0, dropout=0.0):
        super(PWAM, self).__init__()
        # input x shape: (B, H*W, dim)
        self.vis_project = nn.Sequential(nn.Conv1d(dim, dim, 1, 1),  # the init function sets bias to 0 if bias is True
                                         nn.GELU(),
                                         nn.Dropout(dropout)
                                        )

        self.image_lang_att = SpatialImageLanguageAttention(v_in_channels,  # v_in
                                                            l_in_channels,  # l_in
                                                            key_channels,  # key
                                                            value_channels,  # value
                                                            out_channels=value_channels,  # out
                                                            num_heads=num_heads)

        self.project_mm = nn.Sequential(nn.Conv1d(value_channels, value_channels, 1, 1),
                                        nn.GELU(),
                                        nn.Dropout(dropout)
                                        )

    def forward(self, x, l, l_mask):
        # input x shape: (B, H*W, dim)
        vis = self.vis_project(x.permute(0, 2, 1))  # (B, dim, H*W)

        lang = self.image_lang_att(x, l.permute(0,2,1), l_mask)  # (B, H*W, dim)

        lang = lang.permute(0, 2, 1)  # (B, dim, H*W)

        mm = torch.mul(vis, lang)
        mm = self.project_mm(mm)  # (B, dim, H*W)

        mm = mm.permute(0, 2, 1)  # (B, H*W, dim)

        return mm


class SpatialImageLanguageAttention(nn.Module):
    def __init__(self, v_in_channels, l_in_channels, key_channels, value_channels, out_channels=None, num_heads=1):
        super(SpatialImageLanguageAttention, self).__init__()
        # x shape: (B, H*W, v_in_channels)
        # l input shape: (B, l_in_channels, N_l)
        # l_mask shape: (B, N_l, 1)
        self.v_in_channels = v_in_channels
        self.l_in_channels = l_in_channels
        self.out_channels = out_channels
        self.key_channels = key_channels
        self.value_channels = value_channels
        self.num_heads = num_heads
        if out_channels is None:
            self.out_channels = self.value_channels

        # Keys: language features: (B, l_in_channels, #words)
        # avoid any form of spatial normalization because a sentence contains many padding 0s
        self.f_key = nn.Sequential(
            nn.Conv1d(self.l_in_channels, self.key_channels, kernel_size=1, stride=1),
        )

        # Queries: visual features: (B, H*W, v_in_channels)
        self.f_query = nn.Sequential(
            #nn.Conv1d(self.v_in_channels, self.key_channels, kernel_size=1, stride=1),
            #nn.InstanceNorm1d(self.key_channels),
            #nn.Conv1d(self.v_in_channels, self.key_channels, kernel_size=1, stride=1),
            nn.Linear(self.v_in_channels, self.key_channels),
            #nn.InstanceNorm1d(self.key_channels),
            nn.LayerNorm(self.key_channels),
        )

        # Values: language features: (B, l_in_channels, #words)
        self.f_value = nn.Sequential(
            nn.Conv1d(self.l_in_channels, self.value_channels, kernel_size=1, stride=1),
        )

        # Out projection
        self.W = nn.Sequential(
            #nn.Conv1d(self.value_channels, self.out_channels, kernel_size=1, stride=1),
            #nn.InstanceNorm1d(self.out_channels),
            #nn.Conv1d(self.value_channels, self.out_channels, kernel_size=1, stride=1),
            nn.Linear(self.value_channels, self.out_channels),
            #nn.InstanceNorm1d(self.out_channels),
            nn.LayerNorm(self.out_channels),
        )

    def forward(self, x, l, l_mask):
        # x shape: (B, H*W, v_in_channels)
        # l input shape: (B, l_in_channels, N_l)
        # l_mask shape: (B, N_l, 1)
        B, HW = x.size(0), x.size(1)
        #x = x.permute(0, 2, 1)  # (B, key_channels, H*W)
        l_mask = l_mask.permute(0, 2, 1)  # (B, N_l, 1) -> (B, 1, N_l)

        query = self.f_query(x).permute(0,2,1)  # (B, key_channels, H*W) if Conv1D
        query = query.permute(0, 2, 1)  # (B, H*W, key_channels)
        key = self.f_key(l)  # (B, key_channels, N_l)
        value = self.f_value(l)  # (B, self.value_channels, N_l)
        key = key * l_mask  # (B, key_channels, N_l)
        value = value * l_mask  # (B, self.value_channels, N_l)
        n_l = value.size(-1)
        query = query.reshape(B, HW, self.num_heads, self.key_channels//self.num_heads).permute(0, 2, 1, 3)
        # (b, num_heads, H*W, self.key_channels//self.num_heads)
        key = key.reshape(B, self.num_heads, self.key_channels//self.num_heads, n_l)
        # (b, num_heads, self.key_channels//self.num_heads, n_l)
        value = value.reshape(B, self.num_heads, self.value_channels//self.num_heads, n_l)
        # # (b, num_heads, self.value_channels//self.num_heads, n_l)
        l_mask = l_mask.unsqueeze(1)  # (b, 1, 1, n_l)

        sim_map = torch.matmul(query, key)  # (B, self.num_heads, H*W, N_l)
        sim_map = (self.key_channels ** -.5) * sim_map  # scaled dot product

        sim_map = sim_map + (1e4*l_mask - 1e4)  # assign a very small number to padding positions
        sim_map = F.softmax(sim_map, dim=-1)  # (B, num_heads, h*w, N_l)
        out = torch.matmul(sim_map, value.permute(0, 1, 3, 2))  # (B, num_heads, H*W, self.value_channels//num_heads)
        out = out.permute(0, 2, 1, 3).contiguous().reshape(B, HW, self.value_channels)  # (B, H*W, value_channels)
        #out = out.permute(0, 2, 1)  # (B, value_channels, HW)
        #out = self.W(out)  # (B, value_channels, HW)
        #out = out.permute(0, 2, 1)  # (B, HW, value_channels)
        out = self.W(out)

        return out