openrec/modeling/encoders/svtrnet2dpos.py

import numpy as np
import torch
from torch import nn
from torch.nn.init import kaiming_normal_, ones_, trunc_normal_, zeros_

from openrec.modeling.common import DropPath, Identity, Mlp


class ConvBNLayer(nn.Module):

    def __init__(
        self,
        in_channels,
        out_channels,
        kernel_size=3,
        stride=1,
        padding=0,
        bias=False,
        groups=1,
        act=nn.GELU,
    ):
        super().__init__()
        self.conv = nn.Conv2d(
            in_channels=in_channels,
            out_channels=out_channels,
            kernel_size=kernel_size,
            stride=stride,
            padding=padding,
            groups=groups,
            bias=bias,
        )
        self.norm = nn.BatchNorm2d(out_channels)
        self.act = act()

    def forward(self, inputs):
        out = self.conv(inputs)
        out = self.norm(out)
        out = self.act(out)
        return out


class ConvMixer(nn.Module):

    def __init__(
        self,
        dim,
        num_heads=8,
        HW=[8, 25],
        local_k=[3, 3],
    ):
        super().__init__()
        self.HW = HW
        self.dim = dim
        self.local_mixer = nn.Conv2d(dim,
                                     dim,
                                     local_k,
                                     1, [local_k[0] // 2, local_k[1] // 2],
                                     groups=num_heads)

    def forward(self, x, w):
        x = x.transpose(1, 2).reshape([x.shape[0], self.dim, -1, w])
        x = self.local_mixer(x)
        x = x.flatten(2).transpose(1, 2)
        return x


class ConvMlp(nn.Module):

    def __init__(
        self,
        in_features,
        hidden_features=None,
        out_features=None,
        act_layer=nn.GELU,
        drop=0.0,
        groups=1,
    ):
        super().__init__()
        out_features = out_features or in_features
        hidden_features = hidden_features or in_features
        self.fc1 = nn.Conv2d(in_features, hidden_features, 1, groups=groups)
        self.act = act_layer()
        self.fc2 = nn.Conv2d(hidden_features, out_features, 1)
        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


class ConvBlock(nn.Module):

    def __init__(
        self,
        dim,
        num_heads,
        mixer='Global',
        local_mixer=[7, 11],
        HW=None,
        mlp_ratio=4.0,
        qkv_bias=False,
        qk_scale=None,
        drop=0.0,
        attn_drop=0.0,
        drop_path=0.0,
        act_layer=nn.GELU,
        norm_layer='nn.LayerNorm',
        eps=1e-6,
        prenorm=True,
    ):
        super().__init__()

        self.norm1 = nn.BatchNorm2d(dim)
        self.local_mixer = nn.Conv2d(dim,
                                     dim, [5, 5],
                                     1, [2, 2],
                                     groups=num_heads)
        self.drop_path = DropPath(drop_path) if drop_path > 0.0 else Identity()
        self.norm2 = nn.BatchNorm2d(dim)
        mlp_hidden_dim = int(dim * mlp_ratio)
        self.mlp = ConvMlp(in_features=dim,
                           hidden_features=mlp_hidden_dim,
                           act_layer=act_layer,
                           drop=drop)
        self.prenorm = prenorm

    def forward(self, x):
        x = self.norm1(x + self.drop_path(self.local_mixer(x)))
        x = self.norm2(x + self.drop_path(self.mlp(x)))
        return x


class Attention(nn.Module):

    def __init__(
        self,
        dim,
        num_heads=8,
        mixer='Global',
        HW=None,
        local_k=[7, 11],
        qkv_bias=False,
        qk_scale=None,
        attn_drop=0.0,
        proj_drop=0.0,
    ):
        super().__init__()
        self.num_heads = num_heads
        self.dim = dim
        self.head_dim = dim // num_heads
        self.scale = qk_scale or self.head_dim**-0.5

        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)
        self.HW = HW
        if HW is not None:
            H = HW[0]
            W = HW[1]
            if W == -1:
                W = 300
            self.C = dim
            self.H = H
            self.W = W
        if mixer == 'Local' and HW is not None:
            if HW[1] == -1:
                wk = 29
            else:
                wk = local_k[1]
            self.wk = wk
            mask = torch.ones(W, W, dtype=torch.float32, requires_grad=False)

            for w in range(0, W):
                b_w = w - wk // 2 if w - wk // 2 > 0 else 0
                if b_w > W - wk:
                    b_w = W - wk
                mask[w, b_w:b_w + wk] = 0.0
            mask[mask >= 1] = -np.inf

            self.register_buffer('mask', mask)
        self.mixer = mixer

    def forward(self, x, w):
        B, N, _ = x.shape
        h = N // w
        qkv = self.qkv(x).reshape(B, N, 3, self.num_heads,
                                  self.head_dim).permute(2, 0, 3, 1, 4)
        q, k, v = qkv.unbind(0)
        q = q * self.scale
        attn = q @ k.transpose(-2, -1)
        if self.mixer == 'Local' and w >= 32:
            mask1 = self.mask[(self.W - w) // 2:-(self.W - w) // 2,
                              (self.W - w) // 2:-(self.W - w) // 2]
            mask1[:(self.wk // 2 + 1)] = self.mask[:(self.wk // 2 + 1), :w]
            mask1[-(self.wk // 2 + 1):] = self.mask[-(self.wk // 2 + 1):, -w:]
            attn += mask1[None, None, :, :].tile(B, 1, h, h)
        attn = attn.softmax(dim=-1)
        attn = self.attn_drop(attn)
        x = attn @ v

        x = x.transpose(1, 2).reshape(B, N, self.dim)
        x = self.proj(x)
        x = self.proj_drop(x)
        return x


class Block(nn.Module):

    def __init__(
        self,
        dim,
        num_heads,
        mixer='Global',
        local_mixer=[7, 11],
        HW=None,
        mlp_ratio=4.0,
        qkv_bias=False,
        qk_scale=None,
        drop=0.0,
        attn_drop=0.0,
        drop_path=0.0,
        act_layer=nn.GELU,
        norm_layer='nn.LayerNorm',
        eps=1e-6,
    ):
        super().__init__()
        if isinstance(norm_layer, str):
            self.norm1 = eval(norm_layer)(dim, eps=eps)
        else:
            self.norm1 = norm_layer(dim)
        if mixer == 'Global' or mixer == 'Local':
            self.mixer = Attention(
                dim,
                num_heads=num_heads,
                mixer=mixer,
                HW=HW,
                local_k=local_mixer,
                qkv_bias=qkv_bias,
                qk_scale=qk_scale,
                attn_drop=attn_drop,
                proj_drop=drop,
            )
        elif mixer == 'Conv':
            self.mixer = ConvMixer(dim,
                                   num_heads=num_heads,
                                   HW=HW,
                                   local_k=local_mixer)
        else:
            raise TypeError('The mixer must be one of [Global, Local, Conv]')

        self.drop_path = DropPath(drop_path) if drop_path > 0.0 else Identity()
        if isinstance(norm_layer, str):
            self.norm2 = eval(norm_layer)(dim, eps=eps)
        else:
            self.norm2 = norm_layer(dim)
        mlp_hidden_dim = int(dim * mlp_ratio)
        self.mlp_ratio = mlp_ratio
        self.mlp = Mlp(
            in_features=dim,
            hidden_features=mlp_hidden_dim,
            act_layer=act_layer,
            drop=drop,
        )

    def forward(self, x, w):
        x = self.norm1(x + self.drop_path(self.mixer(x, w)))
        x = self.norm2(x + self.drop_path(self.mlp(x)))
        return x, w


class PatchEmbed(nn.Module):
    """Image to Patch Embedding."""

    def __init__(
        self,
        img_size=[32, 100],
        in_channels=3,
        embed_dim=768,
        sub_num=2,
        patch_size=[4, 4],
        mode='pope',
    ):
        super().__init__()
        num_patches = (img_size[1] // (2**sub_num)) * (img_size[0] //
                                                       (2**sub_num))
        self.img_size = img_size
        self.num_patches = num_patches
        self.embed_dim = embed_dim
        self.norm = None
        if mode == 'pope':
            if sub_num == 2:
                self.proj = nn.Sequential(
                    ConvBNLayer(
                        in_channels=in_channels,
                        out_channels=embed_dim // 2,
                        kernel_size=3,
                        stride=2,
                        padding=1,
                        act=nn.GELU,
                        bias=None,
                    ),
                    ConvBNLayer(
                        in_channels=embed_dim // 2,
                        out_channels=embed_dim,
                        kernel_size=3,
                        stride=2,
                        padding=1,
                        act=nn.GELU,
                        bias=None,
                    ),
                )
            if sub_num == 3:
                self.proj = nn.Sequential(
                    ConvBNLayer(
                        in_channels=in_channels,
                        out_channels=embed_dim // 4,
                        kernel_size=3,
                        stride=2,
                        padding=1,
                        act=nn.GELU,
                        bias=None,
                    ),
                    ConvBNLayer(
                        in_channels=embed_dim // 4,
                        out_channels=embed_dim // 2,
                        kernel_size=3,
                        stride=2,
                        padding=1,
                        act=nn.GELU,
                        bias=None,
                    ),
                    ConvBNLayer(
                        in_channels=embed_dim // 2,
                        out_channels=embed_dim,
                        kernel_size=3,
                        stride=2,
                        padding=1,
                        act=nn.GELU,
                        bias=None,
                    ),
                )
        elif mode == 'linear':
            self.proj = nn.Conv2d(1,
                                  embed_dim,
                                  kernel_size=patch_size,
                                  stride=patch_size)
            self.num_patches = img_size[0] // patch_size[0] * img_size[
                1] // patch_size[1]

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


class SubSample(nn.Module):

    def __init__(
        self,
        in_channels,
        out_channels,
        types='Pool',
        stride=[2, 1],
        sub_norm='nn.LayerNorm',
        act=None,
    ):
        super().__init__()
        self.types = types
        if types == 'Pool':
            self.avgpool = nn.AvgPool2d(kernel_size=[3, 5],
                                        stride=stride,
                                        padding=[1, 2])
            self.maxpool = nn.MaxPool2d(kernel_size=[3, 5],
                                        stride=stride,
                                        padding=[1, 2])
            self.proj = nn.Linear(in_channels, out_channels)
        else:
            self.conv = nn.Conv2d(in_channels,
                                  out_channels,
                                  kernel_size=3,
                                  stride=stride,
                                  padding=1)
        self.dim = in_channels
        self.norm = eval(sub_norm)(out_channels)
        if act is not None:
            self.act = act()
        else:
            self.act = None

    def forward(self, x, w):
        if self.types == 'Pool':
            x1 = self.avgpool(x)
            x2 = self.maxpool(x)
            x = (x1 + x2) * 0.5
            out = self.proj(x.flatten(2).transpose(1, 2))
        else:
            x = x.transpose(1, 2).reshape([x.shape[0], self.dim, -1, w])
            x = self.conv(x)
            out = x.flatten(2).transpose(1, 2)
        out = self.norm(out)
        if self.act is not None:
            out = self.act(out)

        return out, w


class FlattenTranspose(nn.Module):

    def forward(self, x):
        return x.flatten(2).transpose(1, 2)


class DownSConv(nn.Module):

    def __init__(self, in_channels, out_channels):
        super().__init__()
        self.conv = nn.Conv2d(in_channels,
                              out_channels,
                              3,
                              stride=[2, 1],
                              padding=1)
        self.norm = nn.LayerNorm(out_channels)

    def forward(self, x, w):
        B, N, C = x.shape
        x = x.transpose(1, 2).reshape(B, C, -1, w)
        x = self.conv(x)
        w = x.shape[-1]
        x = self.norm(x.flatten(2).transpose(1, 2))
        return x, w


class SVTRNet2DPos(nn.Module):

    def __init__(
        self,
        img_size=[32, -1],
        in_channels=3,
        embed_dim=[64, 128, 256],
        depth=[3, 6, 3],
        num_heads=[2, 4, 8],
        mixer=['Local'] * 6 +
        ['Global'] * 6,  # Local atten, Global atten, Conv
        local_mixer=[[7, 11], [7, 11], [7, 11]],
        patch_merging='Conv',  # Conv, Pool, None
        pool_size=[2, 1],
        max_size=[16, 32],
        mlp_ratio=4,
        qkv_bias=True,
        qk_scale=None,
        drop_rate=0.0,
        last_drop=0.1,
        attn_drop_rate=0.0,
        drop_path_rate=0.1,
        norm_layer='nn.LayerNorm',
        eps=1e-6,
        act='nn.GELU',
        last_stage=True,
        sub_num=2,
        use_first_sub=True,
        flatten=False,
        **kwargs,
    ):
        super().__init__()
        self.img_size = img_size
        self.embed_dim = embed_dim
        self.flatten = flatten
        patch_merging = None if patch_merging != 'Conv' and patch_merging != 'Pool' else patch_merging
        self.patch_embed = PatchEmbed(
            img_size=img_size,
            in_channels=in_channels,
            embed_dim=embed_dim[0],
            sub_num=sub_num,
        )
        if img_size[1] == -1:
            self.HW = [img_size[0] // (2**sub_num), -1]
        else:
            self.HW = [
                img_size[0] // (2**sub_num), img_size[1] // (2**sub_num)
            ]
        pos_embed = torch.zeros([1, max_size[0] * max_size[1], embed_dim[0]],
                                dtype=torch.float32)
        trunc_normal_(pos_embed, mean=0, std=0.02)
        self.pos_embed = nn.Parameter(
            pos_embed.transpose(1, 2).reshape(1, embed_dim[0], max_size[0],
                                              max_size[1]),
            requires_grad=True,
        )
        self.pos_drop = nn.Dropout(p=drop_rate)
        conv_block_num = sum(
            [1 if mixer_type == 'ConvB' else 0 for mixer_type in mixer])
        Block_unit = [ConvBlock for _ in range(conv_block_num)
                      ] + [Block for _ in range(len(mixer) - conv_block_num)]
        HW = self.HW
        dpr = np.linspace(0, drop_path_rate, sum(depth))
        self.conv_blocks1 = nn.ModuleList([
            Block_unit[0:depth[0]][i](
                dim=embed_dim[0],
                num_heads=num_heads[0],
                mixer=mixer[0:depth[0]][i],
                HW=self.HW,
                local_mixer=local_mixer[0],
                mlp_ratio=mlp_ratio,
                qkv_bias=qkv_bias,
                qk_scale=qk_scale,
                drop=drop_rate,
                act_layer=eval(act),
                attn_drop=attn_drop_rate,
                drop_path=dpr[0:depth[0]][i],
                norm_layer=norm_layer,
                eps=eps,
            ) for i in range(depth[0])
        ])
        if patch_merging is not None:
            if use_first_sub:
                stride = [2, 1]
                HW = [self.HW[0] // 2, self.HW[1]]
            else:
                stride = [1, 1]
                HW = self.HW

            sub_sample1 = nn.Sequential(
                nn.Conv2d(embed_dim[0],
                          embed_dim[1],
                          3,
                          stride=stride,
                          padding=1),
                nn.BatchNorm2d(embed_dim[1]),
            )
            self.conv_blocks1.append(sub_sample1)

        self.patch_merging = patch_merging
        self.trans_blocks = nn.ModuleList()
        for i in range(depth[1]):
            block = Block_unit[depth[0]:depth[0] + depth[1]][i](
                dim=embed_dim[1],
                num_heads=num_heads[1],
                mixer=mixer[depth[0]:depth[0] + depth[1]][i],
                HW=HW,
                local_mixer=local_mixer[1],
                mlp_ratio=mlp_ratio,
                qkv_bias=qkv_bias,
                qk_scale=qk_scale,
                drop=drop_rate,
                act_layer=eval(act),
                attn_drop=attn_drop_rate,
                drop_path=dpr[depth[0]:depth[0] + depth[1]][i],
                norm_layer=norm_layer,
                eps=eps,
            )
            if i + depth[0] < conv_block_num:
                self.conv_blocks1.append(block)
            else:
                self.trans_blocks.append(block)
        if patch_merging is not None:
            self.trans_blocks.append(DownSConv(embed_dim[1], embed_dim[2]))
            HW = [HW[0] // 2, -1]

        for i in range(depth[2]):
            self.trans_blocks.append(Block_unit[depth[0] + depth[1]:][i](
                dim=embed_dim[2],
                num_heads=num_heads[2],
                mixer=mixer[depth[0] + depth[1]:][i],
                HW=HW,
                local_mixer=local_mixer[2],
                mlp_ratio=mlp_ratio,
                qkv_bias=qkv_bias,
                qk_scale=qk_scale,
                drop=drop_rate,
                act_layer=eval(act),
                attn_drop=attn_drop_rate,
                drop_path=dpr[depth[0] + depth[1]:][i],
                norm_layer=norm_layer,
                eps=eps,
            ))
        self.last_stage = last_stage
        self.out_channels = embed_dim[-1]

        self.apply(self._init_weights)

    def _init_weights(self, m):
        if isinstance(m, nn.Linear):
            trunc_normal_(m.weight, mean=0, std=0.02)
            if isinstance(m, nn.Linear) and m.bias is not None:
                zeros_(m.bias)
        if isinstance(m, nn.LayerNorm):
            zeros_(m.bias)
            ones_(m.weight)
        if isinstance(m, nn.Conv2d):
            kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')

    @torch.jit.ignore
    def no_weight_decay(self):
        return {'pos_embed', 'sub_sample1', 'sub_sample2'}

    def forward(self, x):
        x = self.patch_embed(x)

        w = x.shape[-1]
        x = x + self.pos_embed[:, :, :x.shape[-2], :w]

        for blk in self.conv_blocks1:
            x = blk(x)

        x = x.flatten(2).transpose(1, 2)
        for blk in self.trans_blocks:
            x, w = blk(x, w)
        B, N, C = x.shape
        if not self.flatten:
            x = x.transpose(1, 2).reshape(B, C, -1, w)

        return x