import torch import torch.nn as nn import timm import types import math import torch.nn.functional as F import clip activations = {} def get_activation(name): def hook(model, input, output): activations[name] = output return hook attention = {} def get_attention(name): def hook(module, input, output): x = input[0] B, N, C = x.shape qkv = ( module.qkv(x) .reshape(B, N, 3, module.num_heads, C // module.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) attn = (q @ k.transpose(-2, -1)) * module.scale attn = attn.softmax(dim=-1) # [:,:,1,1:] attention[name] = attn return hook def get_mean_attention_map(attn, token, shape): attn = attn[:, :, token, 1:] attn = attn.unflatten(2, torch.Size([shape[2] // 16, shape[3] // 16])).float() attn = torch.nn.functional.interpolate( attn, size=shape[2:], mode="bicubic", align_corners=False ).squeeze(0) all_attn = torch.mean(attn, 0) return all_attn class Slice(nn.Module): def __init__(self, start_index=1): super(Slice, self).__init__() self.start_index = start_index def forward(self, x): return x[:, self.start_index :] class AddReadout(nn.Module): def __init__(self, start_index=1): super(AddReadout, self).__init__() self.start_index = start_index def forward(self, x): if self.start_index == 2: readout = (x[:, 0] + x[:, 1]) / 2 else: readout = x[:, 0] return x[:, self.start_index :] + readout.unsqueeze(1) class ProjectReadout(nn.Module): def __init__(self, in_features, start_index=1): super(ProjectReadout, self).__init__() self.start_index = start_index self.project = nn.Sequential(nn.Linear(2 * in_features, in_features), nn.GELU()) def forward(self, x): readout = x[:, 0].unsqueeze(1).expand_as(x[:, self.start_index :]) features = torch.cat((x[:, self.start_index :], readout), -1) return self.project(features) class Transpose(nn.Module): def __init__(self, dim0, dim1): super(Transpose, self).__init__() self.dim0 = dim0 self.dim1 = dim1 def forward(self, x): x = x.transpose(self.dim0, self.dim1) return x def forward_vit(pretrained, x): b, c, h, w = x.shape # encoder glob = pretrained.model.forward_flex(x) layer_1 = pretrained.activations["1"] layer_2 = pretrained.activations["2"] layer_3 = pretrained.activations["3"] layer_4 = pretrained.activations["4"] layer_1 = pretrained.act_postprocess1[0:2](layer_1) layer_2 = pretrained.act_postprocess2[0:2](layer_2) layer_3 = pretrained.act_postprocess3[0:2](layer_3) layer_4 = pretrained.act_postprocess4[0:2](layer_4) unflatten = nn.Sequential( nn.Unflatten( 2, torch.Size( [ h // pretrained.model.patch_size[1], w // pretrained.model.patch_size[0], ] ), ) ) if layer_1.ndim == 3: layer_1 = unflatten(layer_1) if layer_2.ndim == 3: layer_2 = unflatten(layer_2) if layer_3.ndim == 3: layer_3 = unflatten(layer_3) if layer_4.ndim == 3: layer_4 = unflatten(layer_4) layer_1 = pretrained.act_postprocess1[3 : len(pretrained.act_postprocess1)](layer_1) layer_2 = pretrained.act_postprocess2[3 : len(pretrained.act_postprocess2)](layer_2) layer_3 = pretrained.act_postprocess3[3 : len(pretrained.act_postprocess3)](layer_3) layer_4 = pretrained.act_postprocess4[3 : len(pretrained.act_postprocess4)](layer_4) return layer_1, layer_2, layer_3, layer_4 def _resize_pos_embed(self, posemb, gs_h, gs_w): posemb_tok, posemb_grid = ( posemb[:, : self.start_index], posemb[0, self.start_index :], ) gs_old = int(math.sqrt(len(posemb_grid))) posemb_grid = posemb_grid.reshape(1, gs_old, gs_old, -1).permute(0, 3, 1, 2) posemb_grid = F.interpolate(posemb_grid, size=(gs_h, gs_w), mode="bilinear") posemb_grid = posemb_grid.permute(0, 2, 3, 1).reshape(1, gs_h * gs_w, -1) posemb = torch.cat([posemb_tok, posemb_grid], dim=1) return posemb def forward_flex(self, x): b, c, h, w = x.shape pos_embed = self._resize_pos_embed( self.pos_embed, h // self.patch_size[1], w // self.patch_size[0] ) B = x.shape[0] if hasattr(self.patch_embed, "backbone"): x = self.patch_embed.backbone(x) if isinstance(x, (list, tuple)): x = x[-1] # last feature if backbone outputs list/tuple of features x = self.patch_embed.proj(x).flatten(2).transpose(1, 2) if getattr(self, "dist_token", None) is not None: cls_tokens = self.cls_token.expand( B, -1, -1 ) # stole cls_tokens impl from Phil Wang, thanks dist_token = self.dist_token.expand(B, -1, -1) x = torch.cat((cls_tokens, dist_token, x), dim=1) else: cls_tokens = self.cls_token.expand( B, -1, -1 ) # stole cls_tokens impl from Phil Wang, thanks x = torch.cat((cls_tokens, x), dim=1) x = x + pos_embed x = self.pos_drop(x) for blk in self.blocks: x = blk(x) x = self.norm(x) return x def get_readout_oper(vit_features, features, use_readout, start_index=1): if use_readout == "ignore": readout_oper = [Slice(start_index)] * len(features) elif use_readout == "add": readout_oper = [AddReadout(start_index)] * len(features) elif use_readout == "project": readout_oper = [ ProjectReadout(vit_features, start_index) for out_feat in features ] else: assert ( False ), "wrong operation for readout token, use_readout can be 'ignore', 'add', or 'project'" return readout_oper def _make_pretrained_clip_vitl16_384( pretrained, use_readout="ignore", hooks=None, enable_attention_hooks=False ): clip_pretrained, _ = clip.load("ViT-B/32", device='cpu', jit=False) model = timm.create_model("vit_large_patch16_384", pretrained=pretrained) hooks = [5, 11, 17, 23] if hooks == None else hooks pretrained = _make_vit_b16_backbone( model, features=[256, 512, 1024, 1024], hooks=hooks, vit_features=1024, use_readout=use_readout, enable_attention_hooks=enable_attention_hooks, ) return clip_pretrained, pretrained def _make_pretrained_clipRN50x16_vitl16_384( pretrained, use_readout="ignore", hooks=None, enable_attention_hooks=False ): clip_pretrained, _ = clip.load("RN50x16", device='cuda', jit=False) model = timm.create_model("vit_large_patch16_384", pretrained=pretrained) hooks = [5, 11, 17, 23] if hooks == None else hooks pretrained = _make_vit_b16_backbone( model, features=[256, 512, 1024, 1024], hooks=hooks, vit_features=1024, use_readout=use_readout, enable_attention_hooks=enable_attention_hooks, ) return clip_pretrained, pretrained def _make_pretrained_clip_vitb32_384(pretrained, use_readout="ignore", hooks=None, enable_attention_hooks=False): clip_pretrained, _ = clip.load("ViT-B/32", device='cuda', jit=False) model = timm.create_model("vit_base_patch32_384", pretrained=pretrained) hooks = [2, 5, 8, 11] if hooks == None else hooks pretrained = _make_vit_b32_backbone( model, features=[96, 192, 384, 768], hooks=hooks, use_readout=use_readout, enable_attention_hooks=False, ) return clip_pretrained, pretrained def _make_vit_b32_backbone( model, features=[96, 192, 384, 768], size=[384, 384], hooks=[2, 5, 8, 11], vit_features=768, use_readout="ignore", start_index=1, enable_attention_hooks=False, ): pretrained = nn.Module() pretrained.model = model pretrained.model.blocks[hooks[0]].register_forward_hook(get_activation("1")) pretrained.model.blocks[hooks[1]].register_forward_hook(get_activation("2")) pretrained.model.blocks[hooks[2]].register_forward_hook(get_activation("3")) pretrained.model.blocks[hooks[3]].register_forward_hook(get_activation("4")) pretrained.activations = activations pretrained.model.patch_size = [32, 32] pretrained.model.start_index = start_index if enable_attention_hooks: pretrained.model.blocks[hooks[0]].attn.register_forward_hook( get_attention("attn_1") ) pretrained.model.blocks[hooks[1]].attn.register_forward_hook( get_attention("attn_2") ) pretrained.model.blocks[hooks[2]].attn.register_forward_hook( get_attention("attn_3") ) pretrained.model.blocks[hooks[3]].attn.register_forward_hook( get_attention("attn_4") ) pretrained.attention = attention readout_oper = get_readout_oper(vit_features, features, use_readout, start_index) pretrained.act_postprocess1 = nn.Sequential( readout_oper[0], Transpose(1, 2), nn.Unflatten(2, torch.Size([size[0] // pretrained.model.patch_size[1], size[1] // pretrained.model.patch_size[0]])), nn.Conv2d( in_channels=vit_features, out_channels=features[0], kernel_size=1, stride=1, padding=0, ), nn.ConvTranspose2d( in_channels=features[0], out_channels=features[0], kernel_size=8, stride=8, padding=0, bias=True, dilation=1, groups=1, ), ) pretrained.act_postprocess2 = nn.Sequential( readout_oper[1], Transpose(1, 2), nn.Unflatten(2, torch.Size([size[0] // pretrained.model.patch_size[1], size[1] // pretrained.model.patch_size[0]])), nn.Conv2d( in_channels=vit_features, out_channels=features[1], kernel_size=1, stride=1, padding=0, ), nn.ConvTranspose2d( in_channels=features[1], out_channels=features[1], kernel_size=4, stride=4, padding=0, bias=True, dilation=1, groups=1, ), ) pretrained.act_postprocess3 = nn.Sequential( readout_oper[2], Transpose(1, 2), nn.Unflatten(2, torch.Size([size[0] // pretrained.model.patch_size[1], size[1] // pretrained.model.patch_size[0]])), nn.Conv2d( in_channels=vit_features, out_channels=features[2], kernel_size=1, stride=1, padding=0, ), nn.ConvTranspose2d( in_channels=features[2], out_channels=features[2], kernel_size=2, stride=2, padding=0, # output_padding=output_padding, bias=True, dilation=1, groups=1, ), ) pretrained.act_postprocess4 = nn.Sequential( readout_oper[3], Transpose(1, 2), nn.Unflatten(2, torch.Size([size[0] // pretrained.model.patch_size[1], size[1] // pretrained.model.patch_size[0]])), nn.Conv2d( in_channels=vit_features, out_channels=features[3], kernel_size=1, stride=1, padding=0, ), ) # We inject this function into the VisionTransformer instances so that # we can use it with interpolated position embeddings without modifying the library source. pretrained.model.forward_flex = types.MethodType(forward_flex, pretrained.model) pretrained.model._resize_pos_embed = types.MethodType( _resize_pos_embed, pretrained.model ) return pretrained def _make_vit_b16_backbone( model, features=[96, 192, 384, 768], size=[384, 384], hooks=[2, 5, 8, 11], vit_features=768, use_readout="ignore", start_index=1, enable_attention_hooks=False, ): pretrained = nn.Module() pretrained.model = model pretrained.model.blocks[hooks[0]].register_forward_hook(get_activation("1")) pretrained.model.blocks[hooks[1]].register_forward_hook(get_activation("2")) pretrained.model.blocks[hooks[2]].register_forward_hook(get_activation("3")) pretrained.model.blocks[hooks[3]].register_forward_hook(get_activation("4")) pretrained.activations = activations if enable_attention_hooks: pretrained.model.blocks[hooks[0]].attn.register_forward_hook( get_attention("attn_1") ) pretrained.model.blocks[hooks[1]].attn.register_forward_hook( get_attention("attn_2") ) pretrained.model.blocks[hooks[2]].attn.register_forward_hook( get_attention("attn_3") ) pretrained.model.blocks[hooks[3]].attn.register_forward_hook( get_attention("attn_4") ) pretrained.attention = attention readout_oper = get_readout_oper(vit_features, features, use_readout, start_index) # 32, 48, 136, 384 pretrained.act_postprocess1 = nn.Sequential( readout_oper[0], Transpose(1, 2), nn.Unflatten(2, torch.Size([size[0] // 16, size[1] // 16])), nn.Conv2d( in_channels=vit_features, out_channels=features[0], kernel_size=1, stride=1, padding=0, ), nn.ConvTranspose2d( in_channels=features[0], out_channels=features[0], kernel_size=4, stride=4, padding=0, bias=True, dilation=1, groups=1, ), ) pretrained.act_postprocess2 = nn.Sequential( readout_oper[1], Transpose(1, 2), nn.Unflatten(2, torch.Size([size[0] // 16, size[1] // 16])), nn.Conv2d( in_channels=vit_features, out_channels=features[1], kernel_size=1, stride=1, padding=0, ), nn.ConvTranspose2d( in_channels=features[1], out_channels=features[1], kernel_size=2, stride=2, padding=0, bias=True, dilation=1, groups=1, ), ) pretrained.act_postprocess3 = nn.Sequential( readout_oper[2], Transpose(1, 2), nn.Unflatten(2, torch.Size([size[0] // 16, size[1] // 16])), nn.Conv2d( in_channels=vit_features, out_channels=features[2], kernel_size=1, stride=1, padding=0, ), ) pretrained.act_postprocess4 = nn.Sequential( readout_oper[3], Transpose(1, 2), nn.Unflatten(2, torch.Size([size[0] // 16, size[1] // 16])), nn.Conv2d( in_channels=vit_features, out_channels=features[3], kernel_size=1, stride=1, padding=0, ), nn.Conv2d( in_channels=features[3], out_channels=features[3], kernel_size=3, stride=2, padding=1, ), ) pretrained.model.start_index = start_index pretrained.model.patch_size = [16, 16] # We inject this function into the VisionTransformer instances so that # we can use it with interpolated position embeddings without modifying the library source. pretrained.model.forward_flex = types.MethodType(forward_flex, pretrained.model) pretrained.model._resize_pos_embed = types.MethodType( _resize_pos_embed, pretrained.model ) return pretrained