# -------------------------------------------------------- # BEIT: BERT Pre-Training of Image Transformers (https://arxiv.org/abs/2106.08254) # Github source: https://github.com/microsoft/unilm/tree/master/beit # Copyright (c) 2021 Microsoft # Licensed under The MIT License [see LICENSE for details] # By Hangbo Bao # Based on timm, mmseg, setr, xcit and swin code bases # https://github.com/rwightman/pytorch-image-models/tree/master/timm # https://github.com/fudan-zvg/SETR # https://github.com/facebookresearch/xcit/ # https://github.com/microsoft/Swin-Transformer # --------------------------------------------------------' import math from functools import partial import torch import torch.nn as nn import torch.nn.functional as F import torch.utils.checkpoint as checkpoint from mmdet.models.builder import BACKBONES from mmdet.utils import get_root_logger from timm.models.layers import drop_path, to_2tuple, trunc_normal_ from ..utils import load_checkpoint class DropPath(nn.Module): """Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks). """ def __init__(self, drop_prob=None): super(DropPath, self).__init__() self.drop_prob = drop_prob def forward(self, x): return drop_path(x, self.drop_prob, self.training) def extra_repr(self): return 'p={}'.format(self.drop_prob) class Mlp(nn.Module): 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.fc2(x) x = self.drop(x) return x class Attention(nn.Module): def __init__(self, dim, num_heads=8, qkv_bias=False, qk_scale=None, attn_drop=0., proj_drop=0., window_size=None, attn_head_dim=None): super().__init__() self.num_heads = num_heads head_dim = dim // num_heads if attn_head_dim is not None: head_dim = attn_head_dim all_head_dim = head_dim * self.num_heads # NOTE scale factor was wrong in my original version, can set manually to be compat with prev weights self.scale = qk_scale or head_dim**-0.5 self.qkv = nn.Linear(dim, all_head_dim * 3, bias=qkv_bias) self.window_size = window_size q_size = window_size[0] rel_sp_dim = 2 * q_size - 1 self.rel_pos_h = nn.Parameter(torch.zeros(rel_sp_dim, head_dim)) self.rel_pos_w = nn.Parameter(torch.zeros(rel_sp_dim, head_dim)) self.attn_drop = nn.Dropout(attn_drop) self.proj = nn.Linear(all_head_dim, dim) self.proj_drop = nn.Dropout(proj_drop) def forward(self, x, H, W, rel_pos_bias=None): B, N, C = x.shape qkv = self.qkv(x) qkv = qkv.reshape(B, N, 3, self.num_heads, -1).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)) attn = calc_rel_pos_spatial(attn, q, self.window_size, self.window_size, self.rel_pos_h, self.rel_pos_w) attn = attn.softmax(dim=-1) attn = self.attn_drop(attn) x = (attn @ v).transpose(1, 2).reshape(B, N, -1) x = self.proj(x) x = self.proj_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 def calc_rel_pos_spatial( attn, q, q_shape, k_shape, rel_pos_h, rel_pos_w, ): """ Spatial Relative Positional Embeddings. """ sp_idx = 0 q_h, q_w = q_shape k_h, k_w = k_shape # Scale up rel pos if shapes for q and k are different. q_h_ratio = max(k_h / q_h, 1.0) k_h_ratio = max(q_h / k_h, 1.0) dist_h = ( torch.arange(q_h)[:, None] * q_h_ratio - torch.arange(k_h)[None, :] * k_h_ratio) dist_h += (k_h - 1) * k_h_ratio q_w_ratio = max(k_w / q_w, 1.0) k_w_ratio = max(q_w / k_w, 1.0) dist_w = ( torch.arange(q_w)[:, None] * q_w_ratio - torch.arange(k_w)[None, :] * k_w_ratio) dist_w += (k_w - 1) * k_w_ratio Rh = rel_pos_h[dist_h.long()] Rw = rel_pos_w[dist_w.long()] B, n_head, q_N, dim = q.shape r_q = q[:, :, sp_idx:].reshape(B, n_head, q_h, q_w, dim) rel_h = torch.einsum('byhwc,hkc->byhwk', r_q, Rh) rel_w = torch.einsum('byhwc,wkc->byhwk', r_q, Rw) attn[:, :, sp_idx:, sp_idx:] = ( attn[:, :, sp_idx:, sp_idx:].view(B, -1, q_h, q_w, k_h, k_w) + rel_h[:, :, :, :, :, None] + rel_w[:, :, :, :, None, :]).view( B, -1, q_h * q_w, k_h * k_w) return attn 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., attn_head_dim=None): 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 q_size = window_size[0] rel_sp_dim = 2 * q_size - 1 self.rel_pos_h = nn.Parameter(torch.zeros(rel_sp_dim, head_dim)) self.rel_pos_w = nn.Parameter(torch.zeros(rel_sp_dim, head_dim)) 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.softmax = nn.Softmax(dim=-1) def forward(self, x, H, W): """ 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 x = x.reshape(B_, H, W, C) pad_l = pad_t = 0 pad_r = (self.window_size[1] - W % self.window_size[1]) % self.window_size[1] pad_b = (self.window_size[0] - H % self.window_size[0]) % self.window_size[0] x = F.pad(x, (0, 0, pad_l, pad_r, pad_t, pad_b)) _, Hp, Wp, _ = x.shape x = window_partition( x, self.window_size[0]) # nW*B, window_size, window_size, C x = x.view(-1, self.window_size[1] * self.window_size[0], C) # nW*B, window_size*window_size, C B_w = x.shape[0] N_w = x.shape[1] qkv = self.qkv(x).reshape(B_w, N_w, 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)) attn = calc_rel_pos_spatial(attn, q, self.window_size, self.window_size, self.rel_pos_h, self.rel_pos_w) attn = self.softmax(attn) attn = self.attn_drop(attn) x = (attn @ v).transpose(1, 2).reshape(B_w, N_w, C) x = self.proj(x) x = self.proj_drop(x) x = x.view(-1, self.window_size[1], self.window_size[0], C) x = window_reverse(x, self.window_size[0], Hp, Wp) # B H' W' C if pad_r > 0 or pad_b > 0: x = x[:, :H, :W, :].contiguous() x = x.view(B_, H * W, C) return x class Block(nn.Module): def __init__(self, dim, num_heads, mlp_ratio=4., qkv_bias=False, qk_scale=None, drop=0., attn_drop=0., drop_path=0., init_values=None, act_layer=nn.GELU, norm_layer=nn.LayerNorm, window_size=None, attn_head_dim=None, window=False): super().__init__() self.norm1 = norm_layer(dim) if not window: self.attn = Attention( dim, num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale, attn_drop=attn_drop, proj_drop=drop, window_size=window_size, attn_head_dim=attn_head_dim) else: self.attn = WindowAttention( dim, num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale, attn_drop=attn_drop, proj_drop=drop, window_size=window_size, attn_head_dim=attn_head_dim) # NOTE: drop path for stochastic depth, we shall see if this is better than dropout here 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) if init_values is not None: self.gamma_1 = nn.Parameter( init_values * torch.ones((dim)), requires_grad=True) self.gamma_2 = nn.Parameter( init_values * torch.ones((dim)), requires_grad=True) else: self.gamma_1, self.gamma_2 = None, None def forward(self, x, H, W): if self.gamma_1 is None: x = x + self.drop_path(self.attn(self.norm1(x), H, W)) x = x + self.drop_path(self.mlp(self.norm2(x))) else: x = x + self.drop_path( self.gamma_1 * self.attn(self.norm1(x), H, W)) x = x + self.drop_path(self.gamma_2 * self.mlp(self.norm2(x))) return x class PatchEmbed(nn.Module): """ Image to Patch Embedding """ def __init__(self, img_size=224, patch_size=16, in_chans=3, embed_dim=768): super().__init__() img_size = to_2tuple(img_size) patch_size = to_2tuple(patch_size) num_patches = (img_size[1] // patch_size[1]) * ( img_size[0] // patch_size[0]) self.patch_shape = (img_size[0] // patch_size[0], img_size[1] // patch_size[1]) self.img_size = img_size self.patch_size = patch_size self.num_patches = num_patches self.proj = nn.Conv2d( in_chans, embed_dim, kernel_size=patch_size, stride=patch_size) def forward(self, x, **kwargs): B, C, H, W = x.shape # FIXME look at relaxing size constraints # assert H == self.img_size[0] and W == self.img_size[1], \ # f"Input image size ({H}*{W}) doesn't match model ({self.img_size[0]}*{self.img_size[1]})." x = self.proj(x) Hp, Wp = x.shape[2], x.shape[3] x = x.flatten(2).transpose(1, 2) return x, (Hp, Wp) class HybridEmbed(nn.Module): """ CNN Feature Map Embedding Extract feature map from CNN, flatten, project to embedding dim. """ def __init__(self, backbone, img_size=224, feature_size=None, in_chans=3, embed_dim=768): super().__init__() assert isinstance(backbone, nn.Module) img_size = to_2tuple(img_size) self.img_size = img_size self.backbone = backbone if feature_size is None: with torch.no_grad(): # FIXME this is hacky, but most reliable way of determining the exact dim of the output feature # map for all networks, the feature metadata has reliable channel and stride info, but using # stride to calc feature dim requires info about padding of each stage that isn't captured. training = backbone.training if training: backbone.eval() o = self.backbone( torch.zeros(1, in_chans, img_size[0], img_size[1]))[-1] feature_size = o.shape[-2:] feature_dim = o.shape[1] backbone.train(training) else: feature_size = to_2tuple(feature_size) feature_dim = self.backbone.feature_info.channels()[-1] self.num_patches = feature_size[0] * feature_size[1] self.proj = nn.Linear(feature_dim, embed_dim) def forward(self, x): x = self.backbone(x)[-1] x = x.flatten(2).transpose(1, 2) x = self.proj(x) return x class Norm2d(nn.Module): def __init__(self, embed_dim): super().__init__() self.ln = nn.LayerNorm(embed_dim, eps=1e-6) def forward(self, x): x = x.permute(0, 2, 3, 1) x = self.ln(x) x = x.permute(0, 3, 1, 2).contiguous() return x @BACKBONES.register_module() class ViT(nn.Module): """ Vision Transformer with support for patch or hybrid CNN input stage """ def __init__(self, img_size=224, patch_size=16, in_chans=3, num_classes=80, embed_dim=768, depth=12, num_heads=12, mlp_ratio=4., qkv_bias=False, qk_scale=None, drop_rate=0., attn_drop_rate=0., drop_path_rate=0., hybrid_backbone=None, norm_layer=None, init_values=None, use_checkpoint=False, use_abs_pos_emb=False, use_rel_pos_bias=False, use_shared_rel_pos_bias=False, out_indices=[11], interval=3, pretrained=None): super().__init__() norm_layer = norm_layer or partial(nn.LayerNorm, eps=1e-6) self.num_classes = num_classes self.num_features = self.embed_dim = embed_dim # num_features for consistency with other models if hybrid_backbone is not None: self.patch_embed = HybridEmbed( hybrid_backbone, img_size=img_size, in_chans=in_chans, embed_dim=embed_dim) else: self.patch_embed = PatchEmbed( img_size=img_size, patch_size=patch_size, in_chans=in_chans, embed_dim=embed_dim) num_patches = self.patch_embed.num_patches self.out_indices = out_indices if use_abs_pos_emb: self.pos_embed = nn.Parameter( torch.zeros(1, num_patches, embed_dim)) else: self.pos_embed = None self.pos_drop = nn.Dropout(p=drop_rate) dpr = [x.item() for x in torch.linspace(0, drop_path_rate, depth) ] # stochastic depth decay rule self.use_rel_pos_bias = use_rel_pos_bias self.use_checkpoint = use_checkpoint self.blocks = nn.ModuleList([ Block( dim=embed_dim, num_heads=num_heads, mlp_ratio=mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dpr[i], norm_layer=norm_layer, init_values=init_values, window_size=(14, 14) if ((i + 1) % interval != 0) else self.patch_embed.patch_shape, window=((i + 1) % interval != 0)) for i in range(depth) ]) if self.pos_embed is not None: trunc_normal_(self.pos_embed, std=.02) self.norm = norm_layer(embed_dim) self.fpn1 = nn.Sequential( nn.ConvTranspose2d(embed_dim, embed_dim, kernel_size=2, stride=2), Norm2d(embed_dim), nn.GELU(), nn.ConvTranspose2d(embed_dim, embed_dim, kernel_size=2, stride=2), ) self.fpn2 = nn.Sequential( nn.ConvTranspose2d(embed_dim, embed_dim, kernel_size=2, stride=2), ) self.fpn3 = nn.Identity() self.fpn4 = nn.MaxPool2d(kernel_size=2, stride=2) self.apply(self._init_weights) self.fix_init_weight() self.pretrained = pretrained def fix_init_weight(self): def rescale(param, layer_id): param.div_(math.sqrt(2.0 * layer_id)) for layer_id, layer in enumerate(self.blocks): rescale(layer.attn.proj.weight.data, layer_id + 1) rescale(layer.mlp.fc2.weight.data, layer_id + 1) def _init_weights(self, 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) def init_weights(self, pretrained=None): """Initialize the weights in backbone. Args: pretrained (str, optional): Path to pre-trained weights. Defaults to None. """ pretrained = pretrained or self.pretrained 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() print(f'load from {pretrained}') load_checkpoint(self, pretrained, strict=False, logger=logger) elif pretrained is None: self.apply(_init_weights) else: raise TypeError('pretrained must be a str or None') def get_num_layers(self): return len(self.blocks) @torch.jit.ignore def no_weight_decay(self): return {'pos_embed', 'cls_token'} def forward_features(self, x): B, C, H, W = x.shape x, (Hp, Wp) = self.patch_embed(x) batch_size, seq_len, _ = x.size() if self.pos_embed is not None: x = x + self.pos_embed x = self.pos_drop(x) features = [] for i, blk in enumerate(self.blocks): if self.use_checkpoint: x = checkpoint.checkpoint(blk, x) else: x = blk(x, Hp, Wp) x = self.norm(x) xp = x.permute(0, 2, 1).reshape(B, -1, Hp, Wp) ops = [self.fpn1, self.fpn2, self.fpn3, self.fpn4] for i in range(len(ops)): features.append(ops[i](xp)) return tuple(features) def forward(self, x): x = self.forward_features(x) return x