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
|
import logging
from typing import List, Optional, Union
import torch
from torch import nn
from torch.nn import Module
logger = logging.getLogger(__name__)
class PositionalEmbedding(Module):
def __init__(self, num_embeddings: int, embedding_dim: int, pad_index: int) -> None:
super().__init__()
self.embedding = nn.Embedding(num_embeddings, embedding_dim, pad_index)
self.pad_index = pad_index
def forward(self, input):
positions = self._make_positions(input, self.pad_index)
return self.embedding(positions)
def max_positions(self):
if self.pad_index is not None:
return self.num_embeddings - self.pad_index - 1
else:
return self.num_embeddings
def _make_positions(self, tensor, pad_index: int):
masked = tensor.ne(pad_index).long()
return torch.cumsum(masked, dim=1) * masked + pad_index
class TransformerEncoderLayer(Module):
def __init__(
self,
embedding_dim: int,
num_attention_heads: int,
ffn_dimension: Optional[int] = None,
dropout: float = 0.1,
normalize_before: bool = False,
scaling: Optional[float] = None,
) -> None:
super().__init__()
# TODO Manually setting scaling is not allowed
ffn_dimension = ffn_dimension or embedding_dim * 4
self.better_transformer = torch.nn.TransformerEncoderLayer(
d_model=embedding_dim,
nhead=num_attention_heads,
dim_feedforward=ffn_dimension,
dropout=dropout,
batch_first=True,
activation="gelu",
norm_first=normalize_before,
)
def _load_from_state_dict(
self, state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs
):
better_to_old_names = {
"better_transformer.self_attn.in_proj_weight": "attention.input_projection.weight",
"better_transformer.self_attn.in_proj_bias": "attention.input_projection.bias",
"better_transformer.self_attn.out_proj.weight": "attention.output_projection.weight",
"better_transformer.self_attn.out_proj.bias": "attention.output_projection.bias",
"better_transformer.linear1.weight": "residual_mlp.mlp.0.weight",
"better_transformer.linear1.bias": "residual_mlp.mlp.0.bias",
"better_transformer.linear2.weight": "residual_mlp.mlp.3.weight",
"better_transformer.linear2.bias": "residual_mlp.mlp.3.bias",
"better_transformer.norm1.weight": "attention_layer_norm.weight",
"better_transformer.norm1.bias": "attention_layer_norm.bias",
"better_transformer.norm2.weight": "final_layer_norm.weight",
"better_transformer.norm2.bias": "final_layer_norm.bias",
}
for better, old in better_to_old_names.items():
better_name = prefix + better
old_name = prefix + old
if old_name in state_dict:
state_dict[better_name] = state_dict[old_name]
state_dict.pop(old_name)
elif better_name in state_dict:
# Do nothing
pass
elif strict:
missing_keys.append(better_name)
super(TransformerEncoderLayer, self)._load_from_state_dict(
state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs
)
def forward(self, input: torch.Tensor, key_padding_mask: torch.Tensor, attn_mask: Optional[torch.Tensor] = None):
# torch.nn.TransformerEncodeLayer's attn_mask and key_padding_mask's
# order is reversed
return self.better_transformer(input.transpose(0, 1), attn_mask, key_padding_mask).transpose(0, 1)
class TransformerEncoder(Module):
def __init__(
self,
vocab_size: int,
embedding_dim: int,
padding_idx: int,
max_seq_len: int,
num_encoder_layers: int,
num_attention_heads: int,
ffn_dimension: Optional[int] = None,
dropout: float = 0.1,
normalize_before: bool = False,
scaling: Optional[float] = None,
return_all_layers: bool = False,
) -> None:
super().__init__()
self.padding_idx = padding_idx
self.token_embedding = nn.Embedding(vocab_size, embedding_dim, padding_idx)
ffn_dimension = ffn_dimension or 4 * embedding_dim
layer = torch.nn.TransformerEncoderLayer(
d_model=embedding_dim,
nhead=num_attention_heads,
dim_feedforward=ffn_dimension,
dropout=dropout,
activation="gelu",
batch_first=True,
norm_first=normalize_before,
)
self.layers = torch.nn.TransformerEncoder(
encoder_layer=layer,
num_layers=num_encoder_layers,
enable_nested_tensor=True,
mask_check=False,
)
self.positional_embedding = PositionalEmbedding(max_seq_len, embedding_dim, padding_idx)
self.embedding_layer_norm = nn.LayerNorm(embedding_dim)
self.dropout = nn.Dropout(dropout)
self.normalize_before = normalize_before
self.return_all_layers = return_all_layers
def forward(
self, tokens: torch.Tensor, attn_mask: Optional[torch.Tensor] = None
) -> Union[torch.Tensor, List[torch.Tensor]]:
if attn_mask is not None:
torch._assert(
attn_mask.is_floating_point() or attn_mask.dtype == torch.bool,
f"Only float or bool types are supported for attn_mask not {attn_mask.dtype}",
)
padding_mask = tokens.eq(self.padding_idx)
token_embeddings = self.token_embedding(tokens)
embedded_positions = self.positional_embedding(tokens)
embedded = token_embeddings + embedded_positions
if not hasattr(self, "normalize_before"):
self.normalize_before = False
if not self.normalize_before:
embedded = self.embedding_layer_norm(embedded)
embedded = self.dropout(embedded)
if self.return_all_layers:
encoded = embedded
# B x T x C
# Then transpose back to T x B x C
states = [encoded.transpose(1, 0)]
for layer in self.layers.layers:
encoded = layer(encoded, src_key_padding_mask=padding_mask, src_mask=attn_mask)
encoded_t = encoded.transpose(1, 0)
states.append(encoded_t)
if self.normalize_before:
for i, state in enumerate(states):
states[i] = self.embedding_layer_norm(state)
return states
else:
# B x T x C
# Then transpose back to T x B x C
encoded = self.layers(embedded, src_key_padding_mask=padding_mask).transpose(1, 0)
if self.normalize_before:
encoded = self.embedding_layer_norm(encoded)
return encoded
def _load_from_state_dict(
self, state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs
):
better_to_old_names = {
"self_attn.in_proj_weight": "attention.input_projection.weight",
"self_attn.in_proj_bias": "attention.input_projection.bias",
"self_attn.out_proj.weight": "attention.output_projection.weight",
"self_attn.out_proj.bias": "attention.output_projection.bias",
"linear1.weight": "residual_mlp.mlp.0.weight",
"linear1.bias": "residual_mlp.mlp.0.bias",
"linear2.weight": "residual_mlp.mlp.3.weight",
"linear2.bias": "residual_mlp.mlp.3.bias",
"norm1.weight": "attention_layer_norm.weight",
"norm1.bias": "attention_layer_norm.bias",
"norm2.weight": "final_layer_norm.weight",
"norm2.bias": "final_layer_norm.bias",
}
for i in range(self.layers.num_layers):
for better, old in better_to_old_names.items():
better_name = prefix + "layers.layers.{}.".format(i) + better
old_name = prefix + "layers.{}.".format(i) + old
if old_name in state_dict:
state_dict[better_name] = state_dict[old_name]
state_dict.pop(old_name)
elif better_name in state_dict:
# Do nothing
pass
elif strict:
missing_keys.append(better_name)
super(TransformerEncoder, self)._load_from_state_dict(
state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs
)
|
