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
|
"""Implementes the attention module for the transformer."""
from typing import Optional, Tuple
import attr
from einops import rearrange
import torch
from torch import einsum
from torch import nn
from torch import Tensor
import torch.nn.functional as F
from text_recognizer.networks.transformer.embeddings.rotary import apply_rotary_pos_emb
@attr.s(eq=False)
class Attention(nn.Module):
"""Standard attention."""
def __attrs_pre_init__(self) -> None:
"""Pre init constructor."""
super().__init__()
dim: int = attr.ib()
num_heads: int = attr.ib()
causal: bool = attr.ib(default=False)
dim_head: int = attr.ib(default=64)
dropout_rate: float = attr.ib(default=0.0)
scale: float = attr.ib(init=False)
dropout: nn.Dropout = attr.ib(init=False)
fc: nn.Linear = attr.ib(init=False)
qkv_fn: nn.Sequential = attr.ib(init=False)
def __attrs_post_init__(self) -> None:
"""Post init configuration."""
self.scale = self.dim ** -0.5
inner_dim = self.dim * self.dim_head
self.query = nn.Linear(self.dim, inner_dim, bias=False)
self.key = nn.Linear(self.dim, inner_dim, bias=False)
self.value = nn.Linear(self.dim, inner_dim, bias=False)
self.dropout = nn.Dropout(p=self.dropout_rate)
# Feedforward
self.fc = nn.Linear(inner_dim, self.dim)
def forward(
self,
x: Tensor,
context: Optional[Tensor] = None,
mask: Optional[Tensor] = None,
context_mask: Optional[Tensor] = None,
rotary_pos_emb: Optional[Tensor] = None,
) -> Tuple[Tensor, Tensor]:
"""Computes the attention."""
b, n, _, device = *x.shape, x.device
q = self.query(x)
k = self.key(context) if context is not None else self.key(x)
v = self.value(context) if context is not None else self.value(x)
q, k, v = map(
lambda t: rearrange(t, "b n (h d) -> b h n d", h=self.num_heads), (q, k, v)
)
q, k, v = (
apply_rotary_emb(q, k, v, rotary_pos_emb)
if rotary_pos_emb is not None and context is None
else (q, k, v,)
)
input_mask = compute_input_mask(b, n, k, mask, context, context_mask, device)
# Compute the attention
energy = einsum("b h i d, b h j d -> b h i j", q, k) * self.scale
mask_value = -torch.finfo(energy.dtype).max
# Apply input mask
if input_mask is not None:
energy = energy.masked_fill_(~input_mask, mask_value)
del input_mask
if self.causal:
energy = apply_causal_mask(energy, mask, mask_value, device)
attn = F.softmax(energy, dim=-1)
attn = self.dropout(attn)
out = einsum("b h i j, b h j d -> b h i d", attn, v)
out = rearrange(out, "b h n d -> b n (h d)")
out = self.fc(out)
return out, attn
def apply_rotary_emb(
q: Tensor, k: Tensor, v: Tensor, rotary_pos_emb: Tensor
) -> Tuple[Tensor, Tensor, Tensor]:
l = rotary_pos_emb.shape[-1]
(ql, qr), (kl, kr), (vl, vr) = map(lambda t: (t[..., :l], t[..., l:]), (q, k, v))
ql, kl, vl = map(lambda t: apply_rotary_pos_emb(t, rotary_pos_emb), (ql, kl, vl))
q, k, v = map(lambda t: torch.cat(t, dim=-1), ((ql, qr), (kl, kr), (vl, vr)))
return q, k, v
def compute_input_mask(
b: int,
n: int,
k: Tensor,
mask: Optional[Tensor],
context: Optional[Tensor],
context_mask: Optional[Tensor],
device: str,
) -> Optional[Tensor]:
if any(x is not None for x in (mask, context_mask)):
q_mask = mask if mask is not None else torch.ones((b, n), device=device).bool()
k_mask = q_mask if context is None else context_mask
k_mask = (
torch.ones((b, k.shape[-2]), device=device).bool()
if k_mask is None
else k_mask
)
q_mask = rearrange(q_mask, "b i -> b () i ()")
k_mask = rearrange(k_mask, "b j -> b () () j")
return q_mask * k_mask
return
def apply_causal_mask(
energy: Tensor, mask: Tensor, mask_value: Tensor, device: str
) -> Tensor:
i, j = energy.shape[-2:]
r = torch.arange(i, device=device)
mask = rearrange(r, "i -> () () i ()") < rearrange(r, "j -> () () () j")
mask = F.pad(mask, (j - i, 0), value=False)
energy.masked_fill_(mask, mask_value)
del mask
return energy
|
