modeling_VPT.py

from functools import partial
import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F
from timm.models.layers import drop_path, to_2tuple, trunc_normal_
from timm.models.registry import register_model
import torch.utils.checkpoint as checkpoint
import math
from functools import partial, reduce
from operator import mul

def _cfg(url='', **kwargs):
    return {
        'url': url,
        'num_classes': 400, 'input_size': (3, 224, 224), 'pool_size': None,
        'crop_pct': .9, 'interpolation': 'bicubic',
        'mean': (0.5, 0.5, 0.5), 'std': (0.5, 0.5, 0.5),
        **kwargs
    }


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) -> str:
        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.drop(x)
        # commit this for the orignal BERT implement 
        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., 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
        self.scale = qk_scale or head_dim ** -0.5

        self.qkv = nn.Linear(dim, all_head_dim * 3, bias=False)
        if qkv_bias:
            self.q_bias = nn.Parameter(torch.zeros(all_head_dim))
            self.v_bias = nn.Parameter(torch.zeros(all_head_dim))
        else:
            self.q_bias = None
            self.v_bias = None

        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, p_k, p_v, scale):
        B, N, C = x.shape
        qkv_bias = None
        if self.q_bias is not None:
            qkv_bias = torch.cat((self.q_bias, torch.zeros_like(self.v_bias, requires_grad=False), self.v_bias))
        # qkv = self.qkv(x).reshape(B, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
        qkv = F.linear(input=x, weight=self.qkv.weight, bias=qkv_bias)
        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)

        # promot_reparameterization
        if scale is not None and p_k is not None and p_v is not None:
            scale_k = scale[0].view(-1, 1)
            scale_v = scale[1].view(-1, 1)
            p_k = p_k * torch.maximum(scale_k, torch.ones_like(scale_k))
            p_v = p_v * torch.maximum(scale_v, torch.ones_like(scale_v))
        
        # concat prompts
        if p_k is not None and p_v is not None: 
            k = torch.cat((k, p_k.expand(B, self.num_heads, -1, -1)), dim=2)
            v = torch.cat((v, p_v.expand(B, self.num_heads, -1, -1)), dim=2)

        q = q * self.scale
        attn = (q @ k.transpose(-2, -1))

        
        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


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,
                 attn_head_dim=None):
        super().__init__()
        self.norm1 = norm_layer(dim)
        self.attn = Attention(
            dim, num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale,
            attn_drop=attn_drop, proj_drop=drop, 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 > 0:
            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, p_k, p_v, scale):
        if self.gamma_1 is None:
            x = x + self.drop_path(self.attn(self.norm1(x), p_k, p_v, scale))
            x = x + self.drop_path(self.mlp(self.norm2(x)))
        else:
            x = x + self.drop_path(self.gamma_1 * self.attn(self.norm1(x), p_k, p_v, scale))
            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, num_frames=16, tubelet_size=2):
        super().__init__()
        img_size = to_2tuple(img_size)
        patch_size = to_2tuple(patch_size)
        self.tubelet_size = int(tubelet_size)
        num_patches = (img_size[1] // patch_size[1]) * (img_size[0] // patch_size[0]) * (num_frames // self.tubelet_size)
        self.img_size = img_size
        self.patch_size = patch_size
        self.num_patches = num_patches
        self.proj = nn.Conv3d(in_channels=in_chans, out_channels=embed_dim, 
                            kernel_size = (self.tubelet_size,  patch_size[0],patch_size[1]), 
                            stride=(self.tubelet_size,  patch_size[0],  patch_size[1]))

    def forward(self, x, **kwargs):
        B, C, T, 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)
        return x
    
# sin-cos position encoding
# https://github.com/jadore801120/attention-is-all-you-need-pytorch/blob/master/transformer/Models.py#L31
def get_sinusoid_encoding_table(n_position, d_hid): 
    ''' Sinusoid position encoding table ''' 
    # TODO: make it with torch instead of numpy 
    def get_position_angle_vec(position): 
        return [position / np.power(10000, 2 * (hid_j // 2) / d_hid) for hid_j in range(d_hid)] 

    sinusoid_table = np.array([get_position_angle_vec(pos_i) for pos_i in range(n_position)]) 
    sinusoid_table[:, 0::2] = np.sin(sinusoid_table[:, 0::2]) # dim 2i 
    sinusoid_table[:, 1::2] = np.cos(sinusoid_table[:, 1::2]) # dim 2i+1 

    return  torch.tensor(sinusoid_table,dtype=torch.float, requires_grad=False).unsqueeze(0) 


class PromptedVisionTransformer(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=1000, 
                 embed_dim=768, 
                 depth=12,
                 num_heads=12, 
                 mlp_ratio=4., 
                 qkv_bias=False, 
                 qk_scale=None, 
                 fc_drop_rate=0., 
                 drop_rate=0., 
                 attn_drop_rate=0.,
                 drop_path_rate=0., 
                 norm_layer=nn.LayerNorm, 
                 init_values=0.,
                 use_learnable_pos_emb=False, 
                 init_scale=0.,
                 all_frames=16,
                 tubelet_size=2,
                 prompt_num_tokens=5,
                 prompt_start=0,
                 prompt_end=11,
                 prompt_dropout=0.0,
                 prompt_reparam=False,
                 prompt_init="random",
                 transfer_type='end2end',
                 use_checkpoint=False,
                 use_mean_pooling=True):
        super().__init__()
        self.num_classes = num_classes
        self.num_features = self.embed_dim = embed_dim  # num_features for consistency with other models
        self.tubelet_size = tubelet_size
        self.patch_embed = PatchEmbed(
            img_size=img_size, patch_size=patch_size, in_chans=in_chans, embed_dim=embed_dim, num_frames=all_frames, tubelet_size=self.tubelet_size)
        num_patches = self.patch_embed.num_patches
        self.use_checkpoint = use_checkpoint
        self.transfer_type = transfer_type

        if use_learnable_pos_emb:
            self.pos_embed = nn.Parameter(torch.zeros(1, num_patches, embed_dim))
        else:
            # sine-cosine positional embeddings is on the way
            self.pos_embed = get_sinusoid_encoding_table(num_patches, embed_dim)

        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.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)
            for i in range(depth)])
        self.norm = nn.Identity() if use_mean_pooling else norm_layer(embed_dim)
        self.head_norm = norm_layer(embed_dim) if use_mean_pooling else None
        self.head_dropout = nn.Dropout(p=fc_drop_rate) if fc_drop_rate > 0 else nn.Identity()
        self.head = nn.Linear(embed_dim, num_classes) if num_classes > 0 else nn.Identity()

        if use_learnable_pos_emb:
            trunc_normal_(self.pos_embed, std=.02)

        trunc_normal_(self.head.weight, std=.02)
        self.apply(self._init_weights)

        self.head.weight.data.mul_(init_scale)
        self.head.bias.data.mul_(init_scale)

        if transfer_type == 'prompt':
            # Prompt Tuning 
            self.num_tokens = prompt_num_tokens
            self.prompt_start = prompt_start
            self.prompt_end = prompt_end
            self.prompt_dropout = prompt_dropout
            self.prompt_init = prompt_init
            self.prompt_dropout = nn.Dropout(self.prompt_dropout)
            self.prompt_reparam = prompt_reparam
        
            assert self.prompt_start >=0 and self.prompt_start < depth, "prompt start should be in range [0, depth-1]"
            assert self.prompt_end >=0 and self.prompt_end < depth, "prompt end should be in range [0, depth-1]"
            assert self.prompt_end >=  self.prompt_start, "prompt end should be larger than (or equal) prompt start"

            if self.prompt_init == "random":
                head_dim = self.embed_dim // num_heads
                val = math.sqrt(6. / float(3 * reduce(mul, self.patch_embed.patch_size, 1) + head_dim))  # noqa
            
                self.prompt_embeddings_k= nn.Parameter(torch.zeros(
                    self.prompt_end-self.prompt_start+1,
                    self.num_tokens,
                    head_dim
                ))

                self.prompt_embeddings_v= nn.Parameter(torch.zeros(
                    self.prompt_end-self.prompt_start+1,
                    self.num_tokens,
                    head_dim
                ))

                if self.prompt_reparam:
                    self.prompt_scale = nn.Parameter(torch.ones(
                        self.prompt_end-self.prompt_start+1,
                        2,
                        self.num_tokens
                    ))
                    print("Using reparameterization trick for prompt tuning")
                else:
                    self.prompt_scale = None

                nn.init.uniform_(
                    self.prompt_embeddings_k.data, -val, val)
                nn.init.uniform_(
                    self.prompt_embeddings_v.data, -val, val)
            else:
                raise ValueError("Other initiation scheme is not supported")
        elif transfer_type == 'linear':
            self.prompt_start = depth # This will avoid going through the prompt block
            self.prompt_end = depth - 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 get_num_layers(self):
        return len(self.blocks)

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

    def get_classifier(self):
        return self.head

    def reset_classifier(self, num_classes, global_pool=''):
        self.num_classes = num_classes
        self.head = nn.Linear(self.embed_dim, num_classes) if num_classes > 0 else nn.Identity()

    def forward_features(self, x):
        x = self.patch_embed(x)
        B, N, _ = x.size()

        if self.pos_embed is not None:
            x = x + self.pos_embed.expand(B, -1, -1).type_as(x).to(x.device).clone().detach()
        
        x = self.pos_drop(x)

        B, N_vis, C = x.shape
        num_layers = len(self.blocks)
        
        for i in range(num_layers):    
            if i >= self.prompt_start and i <= self.prompt_end:
                if self.use_checkpoint:
                    print("checkpoint")
                    x = checkpoint.checkpoint(self.blocks[i](x), x)
                else:
                    x = self.blocks[i](x, 
                    self.prompt_dropout(self.prompt_embeddings_k[i-self.prompt_start]), 
                    self.prompt_dropout(self.prompt_embeddings_v[i-self.prompt_start]), 
                    self.prompt_scale[i-self.prompt_start] if self.prompt_scale is not None else None)
            else:
                if self.use_checkpoint:
                    print("checkpoint")
                    x = checkpoint.checkpoint(self.blocks[i](x), x)
                else:
                    x = self.blocks[i](x, None, None, None)    

        x = self.norm(x)
        if self.head_norm is not None:
            return self.head_norm(x.mean(1))
        else:
            return x[:, 0]

    def forward(self, x):
        x = self.forward_features(x)
        x = self.head(self.head_dropout(x))
        return x


@register_model
def vit_small_patch16_224(pretrained=False, **kwargs):
    model = PromptedVisionTransformer(
        patch_size=16, embed_dim=384, depth=12, num_heads=6, mlp_ratio=4, qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6), **kwargs)
    model.default_cfg = _cfg()
    return model


@register_model
def vit_base_patch16_224(pretrained=False, **kwargs):
    model = PromptedVisionTransformer(
        patch_size=16, embed_dim=768, depth=12, num_heads=12, mlp_ratio=4, qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6), **kwargs)
    model.default_cfg = _cfg()
    return model


@register_model
def vit_base_patch16_384(pretrained=False, **kwargs):
    model = PromptedVisionTransformer(
        img_size=384, patch_size=16, embed_dim=768, depth=12, num_heads=12, mlp_ratio=4, qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6), **kwargs)
    model.default_cfg = _cfg()
    return model


@register_model
def vit_large_patch16_224(pretrained=False, **kwargs):
    model = PromptedVisionTransformer(
        patch_size=16, embed_dim=1024, depth=24, num_heads=16, mlp_ratio=4, qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6), **kwargs)
    model.default_cfg = _cfg()
    return model


@register_model
def vit_large_patch16_384(pretrained=False, **kwargs):
    model = PromptedVisionTransformer(
        img_size=384, patch_size=16, embed_dim=1024, depth=24, num_heads=16, mlp_ratio=4, qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6), **kwargs)
    model.default_cfg = _cfg()
    return model


@register_model
def vit_large_patch16_512(pretrained=False, **kwargs):
    model = PromptedVisionTransformer(
        img_size=512, patch_size=16, embed_dim=1024, depth=24, num_heads=16, mlp_ratio=4, qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6), **kwargs)
    model.default_cfg = _cfg()
    return model


@register_model
def vit_huge_patch16_224(pretrained=False, **kwargs):
    model = PromptedVisionTransformer(
        patch_size=16, embed_dim=1280, depth=32, num_heads=16, mlp_ratio=4, qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6), **kwargs)
    model.default_cfg = _cfg()
    return model