forked from snap-stanford/GreaseLM
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathmodeling_roberta.py
More file actions
308 lines (253 loc) · 13 KB
/
Copy pathmodeling_roberta.py
File metadata and controls
308 lines (253 loc) · 13 KB
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
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
# refer from https://github.com/huggingface/transformers/blob/90cddfa824b4c127a088e263ef53e1365acb7a2b/src/transformers/models/roberta/modeling_roberta.py
from transformers.modeling_roberta import RobertaModel,RobertaEmbeddings
import torch
from torch import nn
class RobertaPoolModel(RobertaModel):
def __init__(self):
super().__init__()
self.embeddings = RobertaEmbeddings()
# 直接合并向量表示
class RobertaPoolEmbeddings(RobertaEmbeddings):
def __init__(self, config):
super().__init__(config = config)
self.word_embeddings = nn.Embedding(config.vocab_size, config.hidden_size, padding_idx=config.pad_token_id)
self.position_embeddings = nn.Embedding(config.max_position_embeddings, config.hidden_size)
self.token_type_embeddings = nn.Embedding(config.type_vocab_size, config.hidden_size)
# self.LayerNorm is not snake-cased to stick with TensorFlow model variable name and be able to load
# any TensorFlow checkpoint file
self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
self.dropout = nn.Dropout(config.hidden_dropout_prob)
# position_ids (1, len position emb) is contiguous in memory and exported when serialized
self.position_embedding_type = getattr(config, "position_embedding_type", "absolute")
self.register_buffer("position_ids", torch.arange(config.max_position_embeddings).expand((1, -1)))
self.register_buffer(
"token_type_ids", torch.zeros(self.position_ids.size(), dtype=torch.long), persistent=False
)
# End copy
self.padding_idx = config.pad_token_id
self.position_embeddings = nn.Embedding(
config.max_position_embeddings, config.hidden_size, padding_idx=self.padding_idx
)
def forward(
self, input_ids=None, token_type_ids=None, position_ids=None, inputs_embeds=None, pool_mask=None,past_key_values_length=0
):
if position_ids is None:
if input_ids is not None:
# Create the position ids from the input token ids. Any padded tokens remain padded.
position_ids = create_position_ids_from_input_ids(input_ids, self.padding_idx, past_key_values_length)
else:
position_ids = self.create_position_ids_from_inputs_embeds(inputs_embeds)
if input_ids is not None:
input_shape = input_ids.size()
else:
input_shape = inputs_embeds.size()[:-1]
seq_length = input_shape[1]
# Setting the token_type_ids to the registered buffer in constructor where it is all zeros, which usually occurs
# when its auto-generated, registered buffer helps users when tracing the model without passing token_type_ids, solves
# issue #5664
if token_type_ids is None:
if hasattr(self, "token_type_ids"):
buffered_token_type_ids = self.token_type_ids[:, :seq_length]
buffered_token_type_ids_expanded = buffered_token_type_ids.expand(input_shape[0], seq_length)
token_type_ids = buffered_token_type_ids_expanded
else:
token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=self.position_ids.device)
if inputs_embeds is None:
inputs_embeds = self.word_embeddings(input_ids)
# print(inputs_embeds.shape)
# print(input_ids.shape)
# print(pool_mask.shape)
inputs_embeds = self.pool_input_embeds(inputs_embeds,pool_mask)
token_type_embeddings = self.token_type_embeddings(token_type_ids)
embeddings = inputs_embeds + token_type_embeddings
if self.position_embedding_type == "absolute":
position_embeddings = self.position_embeddings(position_ids)
embeddings += position_embeddings
embeddings = self.LayerNorm(embeddings)
embeddings = self.dropout(embeddings)
return embeddings
def pool_input_embeds(self,inputs_embeds,pool_mask):
# print(pool_mask)
# print(inputs_embeds)
for bs in range(pool_mask.shape[0]):
row_pool = pool_mask[bs]
start,end = 0,0
for i in range(pool_mask.shape[1]-1):
if row_pool[i]==0 and row_pool[i+1]==1:
start = i+1
elif row_pool[i]==1 and i==0:
start = i
elif row_pool[i+1] ==1 and i == pool_mask.shape[1]-2:
end = i+1
elif row_pool[i]==1 and row_pool[i+1]==0:
end = i
if end != 0:
inputs_embeds[bs][start:end+1] = torch.mean(inputs_embeds[bs][start:end+1],dim=0,keepdim=True)
# print(start,end+1)
start,end = 0,0
# print(inputs_embeds)
return inputs_embeds
def create_position_ids_from_inputs_embeds(self, inputs_embeds):
"""
We are provided embeddings directly. We cannot infer which are padded so just generate sequential position ids.
Args:
inputs_embeds: torch.Tensor
Returns: torch.Tensor
"""
input_shape = inputs_embeds.size()[:-1]
sequence_length = input_shape[1]
position_ids = torch.arange(
self.padding_idx + 1, sequence_length + self.padding_idx + 1, dtype=torch.long, device=inputs_embeds.device
)
return position_ids.unsqueeze(0).expand(input_shape)
#添加实体层的embedding
# class RobertaEntityEmbeddings(RobertaEmbeddings):
# def __init__(self, config):
# super().__init__(config = config)
# self.word_embeddings = nn.Embedding(config.vocab_size, config.hidden_size, padding_idx=config.pad_token_id)
# self.position_embeddings = nn.Embedding(config.max_position_embeddings, config.hidden_size)
# self.token_type_embeddings = nn.Embedding(config.type_vocab_size, config.hidden_size)
# self.entity_type_embeddings = nn.Embedding(2, config.hidden_size)
# # self.LayerNorm is not snake-cased to stick with TensorFlow model variable name and be able to load
# # any TensorFlow checkpoint file
# self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
# self.dropout = nn.Dropout(config.hidden_dropout_prob)
# # position_ids (1, len position emb) is contiguous in memory and exported when serialized
# self.position_embedding_type = getattr(config, "position_embedding_type", "absolute")
# self.register_buffer("position_ids", torch.arange(config.max_position_embeddings).expand((1, -1)))
# self.register_buffer(
# "token_type_ids", torch.zeros(self.position_ids.size(), dtype=torch.long), persistent=False
# )
# # End copy
# self.padding_idx = config.pad_token_id
# self.position_embeddings = nn.Embedding(
# config.max_position_embeddings, config.hidden_size, padding_idx=self.padding_idx
# )
# def forward(
# self, input_ids=None, token_type_ids=None, position_ids=None, inputs_embeds=None, pool_mask=None,past_key_values_length=0
# ):
# if position_ids is None:
# if input_ids is not None:
# # Create the position ids from the input token ids. Any padded tokens remain padded.
# position_ids = create_position_ids_from_input_ids(input_ids, self.padding_idx, past_key_values_length)
# else:
# position_ids = self.create_position_ids_from_inputs_embeds(inputs_embeds)
# if input_ids is not None:
# input_shape = input_ids.size()
# else:
# input_shape = inputs_embeds.size()[:-1]
# seq_length = input_shape[1]
# # Setting the token_type_ids to the registered buffer in constructor where it is all zeros, which usually occurs
# # when its auto-generated, registered buffer helps users when tracing the model without passing token_type_ids, solves
# # issue #5664
# if token_type_ids is None:
# if hasattr(self, "token_type_ids"):
# buffered_token_type_ids = self.token_type_ids[:, :seq_length]
# buffered_token_type_ids_expanded = buffered_token_type_ids.expand(input_shape[0], seq_length)
# token_type_ids = buffered_token_type_ids_expanded
# else:
# token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=self.position_ids.device)
# if inputs_embeds is None:
# inputs_embeds = self.word_embeddings(input_ids)
# # print(inputs_embeds.shape)
# # print(input_ids.shape)
# # print(pool_mask.shape)
# inputs_embeds = self.pool_input_embeds(inputs_embeds,pool_mask)
# token_type_embeddings = self.token_type_embeddings(token_type_ids)
# embeddings = inputs_embeds + token_type_embeddings
# if self.position_embedding_type == "absolute":
# position_embeddings = self.position_embeddings(position_ids)
# embeddings += position_embeddings
# embeddings = self.LayerNorm(embeddings)
# embeddings = self.dropout(embeddings)
# return embeddings
# def pool_input_embeds(self,inputs_embeds,pool_mask):
# # print(pool_mask)
# # print(inputs_embeds)
# for bs in range(pool_mask.shape[0]):
# row_pool = pool_mask[bs]
# start,end = 0,0
# for i in range(pool_mask.shape[1]-1):
# if row_pool[i]==0 and row_pool[i+1]==1:
# start = i+1
# elif row_pool[i]==1 and i==0:
# start = i
# elif row_pool[i+1] ==1 and i == pool_mask.shape[1]-2:
# end = i+1
# elif row_pool[i]==1 and row_pool[i+1]==0:
# end = i
# if end != 0:
# inputs_embeds[bs][start:end+1] = torch.mean(inputs_embeds[bs][start:end+1],dim=0,keepdim=True)
# # print(start,end+1)
# start,end = 0,0
# # print(inputs_embeds)
# return inputs_embeds
# def create_position_ids_from_inputs_embeds(self, inputs_embeds):
# """
# We are provided embeddings directly. We cannot infer which are padded so just generate sequential position ids.
# Args:
# inputs_embeds: torch.Tensor
# Returns: torch.Tensor
# """
# input_shape = inputs_embeds.size()[:-1]
# sequence_length = input_shape[1]
# position_ids = torch.arange(
# self.padding_idx + 1, sequence_length + self.padding_idx + 1, dtype=torch.long, device=inputs_embeds.device
# )
# return position_ids.unsqueeze(0).expand(input_shape)
def create_position_ids_from_input_ids(input_ids, padding_idx, past_key_values_length=0):
"""
Replace non-padding symbols with their position numbers. Position numbers begin at padding_idx+1. Padding symbols
are ignored. This is modified from fairseq's `utils.make_positions`.
Args:
x: torch.Tensor x:
Returns: torch.Tensor
"""
# The series of casts and type-conversions here are carefully balanced to both work with ONNX export and XLA.
mask = input_ids.ne(padding_idx).int()
incremental_indices = (torch.cumsum(mask, dim=1).type_as(mask) + past_key_values_length) * mask
return incremental_indices.long() + padding_idx
def pool_input_embeds(inputs_embeds,pool_mask):
# print(pool_mask)
# print(inputs_embeds)
for bs in range(pool_mask.shape[0]):
row_pool = pool_mask[bs]
start,end = 0,0
for i in range(pool_mask.shape[1]-1):
if row_pool[i]==0 and row_pool[i+1]==1:
start = i+1
elif row_pool[i]==1 and i==0:
start = i
elif row_pool[i+1] ==1 and i == pool_mask.shape[1]-2:
end = i+1
elif row_pool[i]==1 and row_pool[i+1]==0:
end = i
if end != 0:
inputs_embeds[bs][start:end+1] = torch.mean(inputs_embeds[bs][start:end+1],dim=0,keepdim=True)
# print(start,end+1)
start,end = 0,0
# print(inputs_embeds)
return inputs_embeds
if __name__ == "__main__":
bs = 2
seq_len = 5
inputs_embeds = torch.randint(1,10,[bs, seq_len,3], dtype=torch.float)
# pool_mask
pool_mask = torch.zeros([bs, seq_len], dtype=torch.float)
pool_mask[0][0:2] = 1
pool_mask[0][3:5] = 1
pool_mask[-1][2:5] = 1
we = nn.Embedding(3,2)
idx = torch.tensor([0,0,1])
# token_type_embeddings = nn.Embedding(config.type_vocab_size, config.hidden_size)
print(pool_mask)
# pool_input_embeds(inputs_embeds,pool_mask)
lm = nn.LayerNorm(5, eps=1e-12)
print(lm.state_dict().keys())
print("参数gamma shape: ", lm.state_dict()['weight'])
print("参数beta shape: ", lm.state_dict()['bias'])
dpo = nn.Dropout(0.7)
we = lm(pool_mask)
print(we)
we = dpo(we)
print(we)