# -------------------------------------------------------- # Swin Transformer # Copyright (c) 2021 Microsoft # Licensed under The MIT License [see LICENSE for details] # Written by Ze Liu # -------------------------------------------------------- import torch import torch.nn as nn import torch.utils.checkpoint as checkpoint from timm.models.layers import DropPath, to_2tuple, trunc_normal_ from timm.data import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD from timm.models.registry import register_model from torch.hub import load_state_dict_from_url import logging from einops import rearrange import math _logger = logging.getLogger(__name__) def _cfg(url='', **kwargs): return { 'url': url, 'num_classes': 1, 'input_size': (3, 224, 224), 'pool_size': None, 'crop_pct': .9, 'interpolation': 'bicubic', 'mean': IMAGENET_DEFAULT_MEAN, 'std': IMAGENET_DEFAULT_STD, 'first_conv': 'patch_embed.proj', 'classifier': 'head', **kwargs } default_cfgs = { # patch models (my experiments) 'swin_base_in1k_patch4_224': _cfg( url='https://github.com/SwinTransformer/storage/releases/download/v1.0.0/swin_base_patch4_window7_224.pth', ), 'swin_base_patch4_window7_224_22k': _cfg( url='https://github.com/SwinTransformer/storage/releases/download/v1.0.0/swin_base_patch4_window7_224_22k.pth', ) } 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.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 # adjust image size for the pyramid structure (i.e. must be integer of 32) def create_new_image_size(img_size, thumbnail_dim, window_size): h, w = img_size * thumbnail_dim[0], img_size * thumbnail_dim[1] new_h, new_w = h, w dim = 32 * window_size if h % (32 * window_size) != 0: new_h = (h // dim + 1) * dim if w % (32 * window_size) != 0: new_w = (w // dim + 1) * dim return (new_h//thumbnail_dim[0], new_w//thumbnail_dim[1]) class WindowAttention(nn.Module): r""" 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): """ 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 def extra_repr(self) -> str: return f'dim={self.dim}, window_size={self.window_size}, num_heads={self.num_heads}' def flops(self, N): # calculate flops for 1 window with token length of N flops = 0 # qkv = self.qkv(x) flops += N * self.dim * 3 * self.dim # attn = (q @ k.transpose(-2, -1)) flops += self.num_heads * N * (self.dim // self.num_heads) * N # x = (attn @ v) flops += self.num_heads * N * N * (self.dim // self.num_heads) # x = self.proj(x) flops += N * self.dim * self.dim return flops class SwinTransformerBlock(nn.Module): r""" Swin Transformer Block. Args: dim (int): Number of input channels. input_resolution (tuple[int]): Input resulotion. 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, input_resolution, 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, bottleneck=False, use_checkpoint=False): super().__init__() self.dim = dim self.input_resolution = input_resolution self.num_heads = num_heads self.window_size = window_size self.shift_size = shift_size self.mlp_ratio = mlp_ratio self.use_checkpoint = use_checkpoint if min(self.input_resolution) <= self.window_size: # if window size is larger than input resolution, we don't partition windows self.shift_size = 0 self.window_size = min(self.input_resolution) 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) if self.shift_size > 0: # calculate attention mask for SW-MSA H, W = self.input_resolution img_mask = torch.zeros((1, H, W, 1)) # 1 H W 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)) else: attn_mask = None self.register_buffer("attn_mask", attn_mask) def forward_attn(self, x): H, W = self.input_resolution B, L, C = x.shape assert L == H * W, "input feature has wrong size" x = self.norm1(x) x = x.view(B, H, W, C) # cyclic shift if self.shift_size > 0: shifted_x = torch.roll(x, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2)) else: shifted_x = x # 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=self.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, H, W) # 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 x = x.view(B, H * W, C) return x def forward_mlp(self, x): return self.drop_path(self.mlp(self.norm2(x))) def forward(self, x): shortcut = x if self.use_checkpoint: x = checkpoint.checkpoint(self.forward_attn, x) else: x = self.forward_attn(x) x = shortcut + self.drop_path(x) if self.use_checkpoint: x = x + checkpoint.checkpoint(self.forward_mlp, x) else: x = x + self.forward_mlp(x) return x def extra_repr(self) -> str: return f"dim={self.dim}, input_resolution={self.input_resolution}, num_heads={self.num_heads}, " \ f"window_size={self.window_size}, shift_size={self.shift_size}, mlp_ratio={self.mlp_ratio}" def flops(self): flops = 0 H, W = self.input_resolution # norm1 flops += self.dim * H * W # W-MSA/SW-MSA nW = H * W / self.window_size / self.window_size flops += nW * self.attn.flops(self.window_size * self.window_size) # mlp flops += 2 * H * W * self.dim * self.dim * self.mlp_ratio # norm2 flops += self.dim * H * W return flops class PatchMerging(nn.Module): r""" 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, input_resolution, dim, norm_layer=nn.LayerNorm): super().__init__() self.input_resolution = input_resolution self.dim = dim self.reduction = nn.Linear(4 * dim, 2 * dim, bias=False) self.norm = norm_layer(4 * dim) def forward(self, x): """ x: B, H*W, C """ H, W = self.input_resolution B, L, C = x.shape assert L == H * W, "input feature has wrong size" assert H % 2 == 0 and W % 2 == 0, f"x size ({H}*{W}) are not even." x = x.view(B, H, W, C) 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 def extra_repr(self) -> str: return f"input_resolution={self.input_resolution}, dim={self.dim}" def flops(self): H, W = self.input_resolution flops = H * W * self.dim flops += (H // 2) * (W // 2) * 4 * self.dim * 2 * self.dim return flops class BasicLayer(nn.Module): """ A basic Swin Transformer layer for one stage. Args: dim (int): Number of input channels. input_resolution (tuple[int]): Input resolution. depth (int): Number of blocks. num_heads (int): Number of attention heads. window_size (int): Local window size. 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 | 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, input_resolution, depth, num_heads, window_size, 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, bottleneck=False): super().__init__() self.dim = dim self.input_resolution = input_resolution self.depth = depth self.use_checkpoint = use_checkpoint # build blocks self.blocks = nn.ModuleList([ SwinTransformerBlock(dim=dim, input_resolution=input_resolution, 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, bottleneck=bottleneck if i == depth-1 else False, use_checkpoint=use_checkpoint) for i in range(depth)]) # patch merging layer if downsample is not None: self.downsample = downsample(input_resolution, dim=dim, norm_layer=norm_layer) else: self.downsample = None def forward(self, x): for blk in self.blocks: if self.use_checkpoint: x = checkpoint.checkpoint(blk, x) else: x = blk(x) if self.downsample is not None: x = self.downsample(x) return x def extra_repr(self) -> str: return f"dim={self.dim}, input_resolution={self.input_resolution}, depth={self.depth}" def flops(self): flops = 0 for blk in self.blocks: flops += blk.flops() if self.downsample is not None: flops += self.downsample.flops() return flops class PatchEmbed(nn.Module): r""" Image to Patch Embedding Args: img_size (int): Image size. Default: 224. 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, img_size=(224,224), patch_size=4, in_chans=3, embed_dim=96, norm_layer=None): super().__init__() #img_size = to_2tuple(img_size) patch_size = to_2tuple(patch_size) patches_resolution = [img_size[0] // patch_size[0], img_size[1] // patch_size[1]] self.img_size = img_size self.patch_size = patch_size self.patches_resolution = patches_resolution self.num_patches = patches_resolution[0] * patches_resolution[1] 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): 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).flatten(2).transpose(1, 2) # B Ph*Pw C if self.norm is not None: x = self.norm(x) return x def flops(self): Ho, Wo = self.patches_resolution flops = Ho * Wo * self.embed_dim * self.in_chans * (self.patch_size[0] * self.patch_size[1]) if self.norm is not None: flops += Ho * Wo * self.embed_dim return flops class SwinTransformer(nn.Module): r""" Swin Transformer A PyTorch impl of : `Swin Transformer: Hierarchical Vision Transformer using Shifted Windows` - https://arxiv.org/pdf/2103.14030 Args: img_size (int | tuple(int)): Input image size. Default 224 patch_size (int | tuple(int)): Patch size. Default: 4 in_chans (int): Number of input image channels. Default: 3 num_classes (int): Number of classes for classification head. Default: 1000 embed_dim (int): Patch embedding dimension. Default: 96 depths (tuple(int)): Depth of each Swin Transformer layer. num_heads (tuple(int)): Number of attention heads in different layers. 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. Default: None drop_rate (float): Dropout rate. Default: 0 attn_drop_rate (float): Attention dropout rate. Default: 0 drop_path_rate (float): Stochastic depth rate. Default: 0.1 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 use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False """ def __init__(self, duration=8, img_size=224, patch_size=4, in_chans=3, num_classes=1000, 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=0.1, norm_layer=nn.LayerNorm, ape=False, patch_norm=True, use_checkpoint=False, thumbnail_rows=1, bottleneck=False, **kwargs): super().__init__() self.duration = duration self.num_classes = num_classes self.num_layers = len(depths) self.embed_dim = embed_dim self.ape = ape self.patch_norm = patch_norm self.num_features = int(embed_dim * 2 ** (self.num_layers - 1)) self.mlp_ratio = mlp_ratio self.thumbnail_rows = thumbnail_rows self.img_size = img_size self.window_size = [window_size for _ in depths] if not isinstance(window_size, list) else window_size self.image_mode = True self.frame_padding = self.duration % thumbnail_rows if self.image_mode is True else 0 if self.frame_padding != 0: self.frame_padding = self.thumbnail_rows - self.frame_padding self.duration += self.frame_padding # split image into non-overlapping patches if self.image_mode: thumbnail_dim = (thumbnail_rows, self.duration // thumbnail_rows) thumbnail_size = (img_size * thumbnail_dim[0], img_size * thumbnail_dim[1]) else: thumbnail_size = (img_size, img_size) print ('---------------------------------------') print ('duration:', self.duration, 'frame padding:', self.frame_padding, 'image_size:', self.img_size, 'patch_size:', patch_size, 'thumbnail_size:', (thumbnail_rows, self.duration//thumbnail_rows), 'ape:', self.ape) self.patch_embed = PatchEmbed( img_size=(img_size, img_size), patch_size=patch_size, in_chans=in_chans, embed_dim=embed_dim, norm_layer=norm_layer if self.patch_norm else None) num_patches = self.patch_embed.num_patches patches_resolution = self.patch_embed.patches_resolution self.patches_resolution = patches_resolution # absolute position embedding if self.ape: # self.absolute_pos_embed = nn.Parameter(torch.zeros(1, num_patches, embed_dim)) # trunc_normal_(self.absolute_pos_embed, std=.02) self.frame_pos_embed = nn.Parameter(torch.zeros(1, self.duration, embed_dim)) trunc_normal_(self.frame_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), input_resolution=(patches_resolution[0] // (2 ** i_layer), patches_resolution[1] // (2 ** i_layer)), depth=depths[i_layer], num_heads=num_heads[i_layer], window_size=self.window_size[i_layer], mlp_ratio=self.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, bottleneck=bottleneck) self.layers.append(layer) self.norm = norm_layer(self.num_features) self.avgpool = nn.AdaptiveAvgPool1d(1) self.head = nn.Linear(self.num_features, num_classes) if num_classes > 0 else nn.Identity() self.apply(self._init_weights) 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) @torch.jit.ignore def no_weight_decay(self): return {'absolute_pos_embed', 'frame_pos_embed'} @torch.jit.ignore def no_weight_decay_keywords(self): return {'relative_position_bias_table'} def create_thumbnail(self, x): # import pdb;pdb.set_trace() input_size = x.shape[-2:] if input_size != to_2tuple(self.img_size): x = nn.functional.interpolate(x, size=self.img_size,mode='bilinear') x = rearrange(x, 'b (th tw c) h w -> b c (th h) (tw w)', th=self.thumbnail_rows, c=3) return x def pad_frames(self, x): frame_num = self.duration - self.frame_padding x = x.view((-1,3*frame_num)+x.size()[2:]) x_padding = torch.zeros((x.shape[0], 3*self.frame_padding) + x.size()[2:]).cuda() x = torch.cat((x, x_padding), dim=1) assert x.shape[1] == 3 * self.duration, 'frame number %d not the same as adjusted input size %d' % (x.shape[1], 3 * self.duration) return x # need to find a better way to do this, maybe torch.fold? def create_image_pos_embed(self): img_rows, img_cols = self.patches_resolution _, _, T = self.frame_pos_embed.shape rows = img_rows // self.thumbnail_rows cols = img_cols // (self.duration // self.thumbnail_rows) img_pos_embed = torch.zeros(img_rows, img_cols, T).cuda() #print (self.duration, T, img_rows, img_cols, rows, cols) for i in range(self.duration): r_indx = (i // self.thumbnail_rows) * rows c_indx = (i % self.thumbnail_rows) * cols img_pos_embed[r_indx:r_indx+rows,c_indx:c_indx+cols] = self.frame_pos_embed[0, i] #print (r_indx, r_indx+rows, c_indx, c_indx+cols) return img_pos_embed.reshape(-1, T) def forward_features(self, x): # x = rearrange(x, 'b (t c) h w -> b c h (t w)', t=self.duration) # in evaluation, it's Bx(num_crops*num_cips*num_frames*3)xHxW # import pdb;pdb.set_trace() b, t, _, h, w = x.shape x = x.view(b, t*3, h, w) if self.frame_padding > 0: x = self.pad_frames(x) else: x = x.view((-1,3*self.duration)+x.size()[2:]) if self.image_mode: x = self.create_thumbnail(x) x = nn.functional.interpolate(x, size=self.img_size,mode='bilinear') else: x = rearrange(x, 'b (n t c) h w -> (b n t) c h w', t=self.duration, c=3) x = self.patch_embed(x) if self.ape: # x = x + self.absolute_pos_embed img_pos_embed = self.create_image_pos_embed() x = x + img_pos_embed x = self.pos_drop(x) for layer in self.layers: x = layer(x) x = self.norm(x) # B L C x = self.avgpool(x.transpose(1, 2)) # B C 1 x = torch.flatten(x, 1) return x def forward(self, x): x = self.forward_features(x) x = self.head(x) if not self.image_mode: x = x.view(-1, self.duration, self.num_classes) x = torch.mean(x, dim=1) return x def flops(self): flops = 0 flops += self.patch_embed.flops() for i, layer in enumerate(self.layers): flops += layer.flops() flops += self.num_features * self.patches_resolution[0] * self.patches_resolution[1] // (2 ** self.num_layers) flops += self.num_features * self.num_classes return flops def load_pretrained(model, cfg=None, num_classes=1000, in_chans=3, filter_fn=None, img_size=224, num_patches=196, pretrained_window_size=7, pretrained_model="", strict=True): if cfg is None: cfg = getattr(model, 'default_cfg') if cfg is None or 'url' not in cfg or not cfg['url']: _logger.warning("Pretrained model URL is invalid, using random initialization.") return if len(pretrained_model) == 0: # state_dict = model_zoo.load_url(cfg['url'], progress=False, map_location='cpu') # state_dict = load_state_dict_from_url(cfg['url'], progress=False, map_location='cpu') state_dict = torch.load('/mnt/new_nas/yansan/models/swin_base_patch4_window7_224_22k.pth', map_location='cpu') else: try: state_dict = load_state_dict(pretrained_model)['model'] except: state_dict = load_state_dict(pretrained_model) if filter_fn is not None: state_dict = filter_fn(state_dict) if in_chans == 1: conv1_name = cfg['first_conv'] _logger.info('Converting first conv (%s) pretrained weights from 3 to 1 channel' % conv1_name) conv1_weight = state_dict[conv1_name + '.weight'] conv1_type = conv1_weight.dtype conv1_weight = conv1_weight.float() O, I, J, K = conv1_weight.shape if I > 3: assert conv1_weight.shape[1] % 3 == 0 # For models with space2depth stems conv1_weight = conv1_weight.reshape(O, I // 3, 3, J, K) conv1_weight = conv1_weight.sum(dim=2, keepdim=False) else: conv1_weight = conv1_weight.sum(dim=1, keepdim=True) conv1_weight = conv1_weight.to(conv1_type) state_dict[conv1_name + '.weight'] = conv1_weight elif in_chans != 3: conv1_name = cfg['first_conv'] conv1_weight = state_dict[conv1_name + '.weight'] conv1_type = conv1_weight.dtype conv1_weight = conv1_weight.float() O, I, J, K = conv1_weight.shape if I != 3: _logger.warning('Deleting first conv (%s) from pretrained weights.' % conv1_name) del state_dict[conv1_name + '.weight'] strict = False else: _logger.info('Repeating first conv (%s) weights in channel dim.' % conv1_name) repeat = int(math.ceil(in_chans / 3)) conv1_weight = conv1_weight.repeat(1, repeat, 1, 1)[:, :in_chans, :, :] conv1_weight *= (3 / float(in_chans)) conv1_weight = conv1_weight.to(conv1_type) state_dict[conv1_name + '.weight'] = conv1_weight #for key, value in state_dict['model'].items(): # print (key) classifier_name = cfg['classifier'] if num_classes == 1000 and cfg['num_classes'] == 1001: # special case for imagenet trained models with extra background class in pretrained weights classifier_weight = state_dict[classifier_name + '.weight'] state_dict[classifier_name + '.weight'] = classifier_weight[1:] classifier_bias = state_dict[classifier_name + '.bias'] state_dict[classifier_name + '.bias'] = classifier_bias[1:] elif num_classes != cfg['num_classes']: # and len(pretrained_model) == 0: # completely discard fully connected for all other differences between pretrained and created model del state_dict['model'][classifier_name + '.weight'] del state_dict['model'][classifier_name + '.bias'] strict = False ''' ## Resizing the positional embeddings in case they don't match if img_size != cfg['input_size'][1]: pos_embed = state_dict['pos_embed'] cls_pos_embed = pos_embed[0, 0, :].unsqueeze(0).unsqueeze(1) other_pos_embed = pos_embed[0, 1:, :].unsqueeze(0).transpose(1, 2) new_pos_embed = F.interpolate(other_pos_embed, size=(num_patches), mode='nearest') new_pos_embed = new_pos_embed.transpose(1, 2) new_pos_embed = torch.cat((cls_pos_embed, new_pos_embed), 1) state_dict['pos_embed'] = new_pos_embed ''' # remove window_size related parameters window_size = (model.window_size)[0] print (pretrained_window_size, window_size) new_state_dict = state_dict['model'].copy() for key in state_dict['model']: if 'attn_mask' in key: del new_state_dict[key] #if window_size != pretrained_window_size: if 1: if 'relative_position_index' in key: del new_state_dict[key] # resize it if 'relative_position_bias_table' in key: #print ('resizing relative_position_bias_table') pretrained_table = state_dict['model'][key] pretrained_table_size = int(math.sqrt(pretrained_table.shape[0])) table_size = int(math.sqrt(model.state_dict()[key].shape[0])) #print (pretrained_table_size, table_size) if pretrained_table_size != table_size: table = pretrained_table.permute(1, 0).view(1, -1, pretrained_table_size, pretrained_table_size) table = nn.functional.interpolate(table, size=table_size, mode='bilinear') table = table.view(-1, table_size*table_size).permute(1, 0) new_state_dict[key] = table for key in model.state_dict(): if 'bottleneck_norm' in key: attn_key = key.replace('bottleneck_norm','norm1') #print (key, attn_key) new_state_dict[key] = new_state_dict[attn_key] ''' for key in new_state_dict: if key not in model.state_dict(): print ('----', key) else: print ('++++', key) print ('====================') for key in model.state_dict(): if key not in new_state_dict: print ('----', key) else: print ('++++', key) ''' print ('loading weights....') ## Loading the weights model_dict = model.state_dict() pretrained_dict = {k: v for k, v in new_state_dict.items() if k in model_dict and model_dict[k].size() == v.size()} model_dict.update(pretrained_dict) model.load_state_dict(model_dict, strict=False) def _conv_filter(state_dict, patch_size=4): """ convert patch embedding weight from manual patchify + linear proj to conv""" out_dict = {} for k, v in state_dict.items(): if 'patch_embed.proj.weight' in k: if v.shape[-1] != patch_size: patch_size = v.shape[-1] v = v.reshape((v.shape[0], 3, patch_size, patch_size)) out_dict[k] = v return out_dict def _create_vision_transformer(variant, pretrained=False, pretrained_window_size=7, **kwargs): default_cfg = default_cfgs[variant] default_num_classes = default_cfg['num_classes'] default_img_size = default_cfg['input_size'][-1] num_classes = kwargs.pop('num_classes', default_num_classes) img_size = kwargs.pop('img_size', default_img_size) repr_size = kwargs.pop('representation_size', None) model_cls = SwinTransformer model = model_cls(img_size=img_size, num_classes=num_classes, **kwargs) model.default_cfg = default_cfg if pretrained: load_pretrained( model, num_classes=num_classes, in_chans=kwargs.get('in_chans', 3), filter_fn=_conv_filter, img_size=img_size, pretrained_window_size=pretrained_window_size, pretrained_model='' ) return model @register_model def TALL_SWIN(pretrained=False, **kwargs): """ ViT-Base (ViT-B/16) from original paper (https://arxiv.org/abs/2010.11929). ImageNet-1k weights fine-tuned from in21k @ 224x224, source https://github.com/google-research/vision_transformer. """ temporal_module_name = kwargs.pop('temporal_module_name', None) temporal_attention_only = kwargs.pop('temporal_attention_only', None) temporal_heads_scale = kwargs.pop('temporal_heads_scale', 1.0) temporal_mlp_scale = kwargs.pop('temporal_mlp_scale', 1.0) rel_pos = kwargs.pop('rel_pos', False) token_maks = kwargs.pop('token_mask', False) frame_cls_tokens = kwargs.pop('frame_cls_tokens', 1) kwargs.pop('hub_attention', '') kwargs.pop('hub_aggregation', '') kwargs.pop('spatial_hub_size', (-1, -1)) kwargs.pop('temporal_pooling', None) kwargs.pop('window_size', -1) embed_dim = 128 mlp_ratio = 4. #drop_path_rate=0.5 patch_size=4 window_size=[14,14,14,7] depths = [2, 2, 18, 2] num_heads = [4, 8, 16, 32] use_checkpoint=kwargs.pop('use_checkpoint', False) ape = kwargs.pop('hpe_to_token', False) bottleneck = True if kwargs.pop('bottleneck', None) is not None else False model_kwargs = dict(patch_size=patch_size, window_size=window_size, embed_dim=embed_dim, depths=depths, num_heads=num_heads, mlp_ratio=mlp_ratio, use_checkpoint=use_checkpoint, ape=ape, bottleneck=bottleneck, **kwargs) print(model_kwargs) model = _create_vision_transformer('swin_base_patch4_window7_224_22k', pretrained=pretrained, pretrained_window_size=7, **model_kwargs) return model if __name__ == '__main__': dummy_input = torch.randn(4,8,3,224,224) model = TALL_SWIN(pretrained=True) print(model(dummy_input))