File size: 8,185 Bytes
9e14838 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 | import torch
from torch import nn, einsum
import torch.nn.functional as F
from einops import rearrange, repeat
from models.efficientnet.efficientnet_pytorch import EfficientNet
# helpers
def exists(val):
return val is not None
# classes
class PreNorm(nn.Module):
def __init__(self, dim, fn):
super().__init__()
self.fn = fn
self.norm = nn.LayerNorm(dim)
def forward(self, x, *args, **kwargs):
x = self.norm(x)
return self.fn(x, *args, **kwargs)
# time token shift
def shift(t, amt):
if amt is 0:
return t
return F.pad(t, (0, 0, 0, 0, amt, -amt))
class PreTokenShift(nn.Module):
def __init__(self, frames, fn):
super().__init__()
self.frames = frames
self.fn = fn
def forward(self, x, *args, **kwargs):
f, dim = self.frames, x.shape[-1]
cls_x, x = x[:, :1], x[:, 1:]
x = rearrange(x, 'b (f n) d -> b f n d', f = f)
# shift along time frame before and after
dim_chunk = (dim // 3)
chunks = x.split(dim_chunk, dim = -1)
chunks_to_shift, rest = chunks[:3], chunks[3:]
shifted_chunks = tuple(map(lambda args: shift(*args), zip(chunks_to_shift, (-1, 0, 1))))
x = torch.cat((*shifted_chunks, *rest), dim = -1)
x = rearrange(x, 'b f n d -> b (f n) d')
x = torch.cat((cls_x, x), dim = 1)
return self.fn(x, *args, **kwargs)
# feedforward
class GEGLU(nn.Module):
def forward(self, x):
x, gates = x.chunk(2, dim = -1)
return x * F.gelu(gates)
class FeedForward(nn.Module):
def __init__(self, dim, mult = 4, dropout = 0.):
super().__init__()
self.net = nn.Sequential(
nn.Linear(dim, dim * mult * 2),
GEGLU(),
nn.Dropout(dropout),
nn.Linear(dim * mult, dim)
)
def forward(self, x):
return self.net(x)
# attention
def attn(q, k, v, mask = None):
sim = einsum('b i d, b j d -> b i j', q, k)
if exists(mask):
max_neg_value = -torch.finfo(sim.dtype).max
sim.masked_fill_(~mask, max_neg_value)
attn = sim.softmax(dim = -1)
out = einsum('b i j, b j d -> b i d', attn, v)
return out
class Attention(nn.Module):
def __init__(
self,
dim,
dim_head = 64,
heads = 8,
dropout = 0.
):
super().__init__()
self.heads = heads
self.scale = dim_head ** -0.5
inner_dim = dim_head * heads
self.to_qkv = nn.Linear(dim, inner_dim * 3, bias = False)
self.to_out = nn.Sequential(
nn.Linear(inner_dim, dim),
nn.Dropout(dropout)
)
def forward(self, x, einops_from, einops_to, mask = None, cls_mask = None, rot_emb = None, **einops_dims):
h = self.heads
q, k, v = self.to_qkv(x).chunk(3, dim = -1)
q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> (b h) n d', h = h), (q, k, v))
q = q * self.scale
# splice out classification token at index 1
(cls_q, q_), (cls_k, k_), (cls_v, v_) = map(lambda t: (t[:, :1], t[:, 1:]), (q, k, v))
# let classification token attend to key / values of all patches across time and space
cls_out = attn(cls_q, k, v, mask = cls_mask)
# rearrange across time or space
q_, k_, v_ = map(lambda t: rearrange(t, f'{einops_from} -> {einops_to}', **einops_dims), (q_, k_, v_))
# expand cls token keys and values across time or space and concat
r = q_.shape[0] // cls_k.shape[0]
cls_k, cls_v = map(lambda t: repeat(t, 'b () d -> (b r) () d', r = r), (cls_k, cls_v))
k_ = torch.cat((cls_k, k_), dim = 1)
v_ = torch.cat((cls_v, v_), dim = 1)
# attention
out = attn(q_, k_, v_, mask = mask)
# merge back time or space
out = rearrange(out, f'{einops_to} -> {einops_from}', **einops_dims)
# concat back the cls token
out = torch.cat((cls_out, out), dim = 1)
# merge back the heads
out = rearrange(out, '(b h) n d -> b n (h d)', h = h)
# combine heads out
return self.to_out(out)
# main classes
class ConvolutionalTimeSformer(nn.Module):
def __init__(
self,
*,
config
):
super().__init__()
self.dim = config['model']['dim']
self.num_frames = config['model']['num-frames']
self.num_patches = config['model']['num-patches']
self.image_size = config['model']['image-size']
self.num_classes = config['model']['num-classes']
self.patch_size = config['model']['patch-size']
self.channels = config['model']['channels']
self.depth = config['model']['depth']
self.heads = config['model']['heads']
self.dim_head = config['model']['dim-head']
self.attn_dropout = config['model']['attn-dropout']
self.ff_dropout = config['model']['ff-dropout']
self.shift_tokens = config['model']['shift-tokens']
self.efficient_net_block = config['model']['efficient-net-block']
self.efficient_net = EfficientNet.from_pretrained('efficientnet-b0')
for m in self.efficient_net.modules():
m.requires_grad = False
self.efficient_net.eval()
num_positions = self.num_frames * self.num_patches
patch_dim = self.patch_size ** 2
self.to_patch_embedding = nn.Linear(patch_dim, self.dim)
self.cls_token = nn.Parameter(torch.randn(1, self.dim))
self.pos_emb = nn.Embedding(num_positions + 1, self.dim)
self.size_emb = nn.Embedding(num_positions + 1, self.dim)
self.layers = nn.ModuleList([])
for _ in range(self.depth):
ff = FeedForward(self.dim, dropout = self.ff_dropout)
time_attn = Attention(self.dim, dim_head = self.dim_head, heads = self.heads, dropout = self.attn_dropout)
spatial_attn = Attention(self.dim, dim_head = self.dim_head, heads = self.heads, dropout = self.attn_dropout)
if self.shift_tokens:
time_attn, spatial_attn, ff = map(lambda t: PreTokenShift(num_frames, t), (time_attn, spatial_attn, ff))
time_attn, spatial_attn, ff = map(lambda t: PreNorm(self.dim, t), (time_attn, spatial_attn, ff))
self.layers.append(nn.ModuleList([time_attn, spatial_attn, ff]))
self.to_out = nn.Sequential(
nn.LayerNorm(self.dim),
nn.Linear(self.dim, self.num_classes)
)
def forward(self, x, mask = None, size_embedding = None):
b, f, h, w, _, *_, device, p = *x.shape, x.device, self.patch_size
hp, wp = (h // p), (w // p)
n = hp * wp
x = rearrange(x, 'b f h w c -> (b f) c h w')
x = self.efficient_net.extract_features_at_block(x, self.efficient_net_block)
x = rearrange(x, '(b f) c h w -> b f c h w', b = b, f = f)
x = rearrange(x, 'b f c h w -> b (f c) (h w)')
tokens = self.to_patch_embedding(x)
# add cls token
cls_token = repeat(self.cls_token, 'n d -> b n d', b = b)
x = torch.cat((cls_token, tokens), dim = 1)
# positional embedding
x += self.pos_emb(torch.arange(x.shape[1], device = device))
# size embedding
size_embedding = repeat(size_embedding, 'b f -> p b f', p=self.num_patches)
size_embedding = rearrange(size_embedding, 'p b f -> (p b f)')
size_embedding = torch.cat((torch.tensor([0]), size_embedding), dim = 0)
size_embedding = size_embedding.to(device)
x += self.size_emb(size_embedding)
# calculate masking for uneven number of frames
frame_mask = None
cls_attn_mask = None
# time and space attention
for (time_attn, spatial_attn, ff) in self.layers:
x = time_attn(x, 'b (f n) d', '(b n) f d', n = n, mask = frame_mask, cls_mask = cls_attn_mask) + x
x = spatial_attn(x, 'b (f n) d', '(b f) n d', f = f, cls_mask = cls_attn_mask) + x
x = ff(x) + x
cls_token = x[:, 0]
return self.to_out(cls_token)
|