ViT源代码学习

import torch
from torch import nn, einsum
import torch.nn.functional as F
from einops import rearrange, repeat
from einops.layers.torch import Rearrange

# helpers
def pair(t):
    return t if isinstance(t, tuple) else (t, t)

# classes
class PreNorm(nn.Module):
    def __init__(self, dim, fn):
        super().__init__()
        self.norm = nn.LayerNorm(dim)
        self.fn = fn
    def forward(self, x, **kwargs):
        return self.fn(self.norm(x), **kwargs)

FFN层

class FeedForward(nn.Module):
    def __init__(self, dim, hidden_dim, dropout = 0.):
        super().__init__()
        self.net = nn.Sequential(
            nn.Linear(dim, hidden_dim),
            nn.GELU(),
            nn.Dropout(dropout),
            nn.Linear(hidden_dim, dim),
            nn.Dropout(dropout)
        )
    def forward(self, x):
        return self.net(x)

x = torch.rand(1, 256, 256, 3)
model = FeedForward(dim=3, hidden_dim=9)
y = model(x)
y.shape

torch.Size([1, 256, 256, 3])

Attention

class Attention(nn.Module):
    def __init__(self, dim, heads = 8, dim_head = 64, dropout = 0.):
        super().__init__()
        inner_dim = dim_head *  heads
        project_out = not (heads == 1 and dim_head == dim)

        self.heads = heads
        self.scale = dim_head ** -0.5

        self.attend = nn.Softmax(dim = -1)
        self.to_qkv = nn.Linear(dim, inner_dim * 3, bias = False)

        self.to_out = nn.Sequential(
            nn.Linear(inner_dim, dim),
            nn.Dropout(dropout)
        ) if project_out else nn.Identity()

    def forward(self, x):
        b, n, _, h = *x.shape, self.heads
        tmp = self.to_qkv(x)
        # print(tmp.shape)
        qkv = self.to_qkv(x).chunk(3, dim=-1)
        # print(qkv[0].shape)
        # print(qkv[1].shape)
        # print(qkv[2].shape)
        q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> b h n d', h = h), qkv)
        # print(q.shape)
        # print(k.shape)
        dots = einsum('b h i d, b h j d -> b h i j', q, k) * self.scale
        # print(dots.shape)
        attn = self.attend(dots)
        # print(attn.shape)
        out = einsum('b h i j, b h j d -> b h i d', attn, v)
        # print(out.shape)
        out = rearrange(out, 'b h n d -> b n (h d)')
        # print(out.shape)
        return self.to_out(out)

class Transformer(nn.Module):
    def __init__(self, dim, depth, heads, dim_head, mlp_dim, dropout = 0.):
        super().__init__()
        self.layers = nn.ModuleList([])
        for _ in range(depth):
            self.layers.append(nn.ModuleList([
                PreNorm(dim, Attention(dim, heads = heads, dim_head = dim_head, dropout = dropout)),
                PreNorm(dim, FeedForward(dim, mlp_dim, dropout = dropout))
            ]))
    def forward(self, x):
        for attn, ff in self.layers:
            x = attn(x) + x
            x = ff(x) + x
        return x

VIT

class ViT(nn.Module):
    def __init__(self, *, image_size, patch_size, num_classes, dim, depth, heads, mlp_dim, pool = 'cls', channels = 3, dim_head = 64, dropout = 0., emb_dropout = 0.):
        super().__init__()
        image_height, image_width = pair(image_size)
        patch_height, patch_width = pair(patch_size)

        assert image_height % patch_height == 0 and image_width % patch_width == 0, 'Image dimensions must be divisible by the patch size.'

        num_patches = (image_height // patch_height) * (image_width // patch_width)
        patch_dim = channels * patch_height * patch_width
        assert pool in {'cls', 'mean'}, 'pool type must be either cls (cls token) or mean (mean pooling)'

        self.to_patch_embedding = nn.Sequential(
            Rearrange('b c (h p1) (w p2) -> b (h w) (p1 p2 c)', p1 = patch_height, p2 = patch_width),
            nn.Linear(patch_dim, dim),
        )

        self.pos_embedding = nn.Parameter(torch.randn(1, num_patches + 1, dim))
        self.cls_token = nn.Parameter(torch.randn(1, 1, dim))
        self.dropout = nn.Dropout(emb_dropout)

        self.transformer = Transformer(dim, depth, heads, dim_head, mlp_dim, dropout)

        self.pool = pool
        self.to_latent = nn.Identity()

        self.mlp_head = nn.Sequential(
            nn.LayerNorm(dim),
            nn.Linear(dim, num_classes)
        )

    def forward(self, img):
        print('input_size: ', img.shape)
        x = self.to_patch_embedding(img)
        print('after patch_embedding: ', x.shape)
        b, n, _ = x.shape
        print('cls_token_size: ', x.shape)
        cls_tokens = repeat(self.cls_token, '() n d -> b n d', b = b)
        x = torch.cat((cls_tokens, x), dim=1)
        print('after cat: ', x.shape)
        x += self.pos_embedding[:, :(n + 1)]
        x = self.dropout(x)
        print('before transformer: ', x.shape)
        x = self.transformer(x)
        print('after transformer: ', x.shape)
        x = x.mean(dim = 1) if self.pool == 'mean' else x[:, 0]
        print(x.shape)
        x = self.to_latent(x)
        print(x.shape)
        return self.mlp_head(x)

x = torch.rand(1, 3, 256, 256)
model = ViT(
            image_size=(256, 256), 
            patch_size=(8, 8),
            num_classes=8,
            dim=128,
            depth=6,
            heads=8,
            mlp_dim=128,    
                )
y = model(x)
y.shape

input_size:  torch.Size([1, 3, 256, 256])
after patch_embedding:  torch.Size([1, 1024, 128])
cls_token_size:  torch.Size([1, 1024, 128])
after cat:  torch.Size([1, 1025, 128])
before transformer:  torch.Size([1, 1025, 128])
after transformer:  torch.Size([1, 1025, 128])
torch.Size([1, 128])
torch.Size([1, 128])

torch.Size([1, 8])