RichardWang
commited on
Commit
•
6549220
1
Parent(s):
aa6806f
add model
Browse files- config.json +4 -2
- modeling_tsp.py +76 -49
- pytorch_model.bin +2 -2
config.json
CHANGED
@@ -1,6 +1,6 @@
|
|
1 |
{
|
2 |
"architectures": [
|
3 |
-
"
|
4 |
],
|
5 |
"auto_map": {
|
6 |
"AutoConfig": "configuration_tsp.TSPConfig",
|
@@ -11,6 +11,7 @@
|
|
11 |
"AutoModelForTokenClassification": "modeling_tsp.TSPModelForTokenClassification"
|
12 |
},
|
13 |
"dropout_prob": 0.1,
|
|
|
14 |
"embedding_size": 128,
|
15 |
"hidden_size": 256,
|
16 |
"intermediate_size": 1024,
|
@@ -21,6 +22,7 @@
|
|
21 |
"pad_token_id": 0,
|
22 |
"position_embedding_type": "absolute",
|
23 |
"torch_dtype": "float32",
|
24 |
-
"transformers_version": "4.
|
|
|
25 |
"vocab_size": 30522
|
26 |
}
|
|
|
1 |
{
|
2 |
"architectures": [
|
3 |
+
"TSPModelForPreTraining"
|
4 |
],
|
5 |
"auto_map": {
|
6 |
"AutoConfig": "configuration_tsp.TSPConfig",
|
|
|
11 |
"AutoModelForTokenClassification": "modeling_tsp.TSPModelForTokenClassification"
|
12 |
},
|
13 |
"dropout_prob": 0.1,
|
14 |
+
"electra_generator_size_divisor": 4,
|
15 |
"embedding_size": 128,
|
16 |
"hidden_size": 256,
|
17 |
"intermediate_size": 1024,
|
|
|
22 |
"pad_token_id": 0,
|
23 |
"position_embedding_type": "absolute",
|
24 |
"torch_dtype": "float32",
|
25 |
+
"transformers_version": "4.19.0.dev0",
|
26 |
+
"use_electra": true,
|
27 |
"vocab_size": 30522
|
28 |
}
|
modeling_tsp.py
CHANGED
@@ -9,12 +9,12 @@ import torch
|
|
9 |
from torch import nn
|
10 |
import torch.nn.functional as F
|
11 |
from transformers import PreTrainedModel
|
12 |
-
from
|
13 |
|
14 |
|
15 |
class TSPPreTrainedModel(PreTrainedModel):
|
16 |
config_class = TSPConfig
|
17 |
-
base_model_prefix = "
|
18 |
|
19 |
def _init_weights(self, module):
|
20 |
"""Initialize the weights"""
|
@@ -32,20 +32,21 @@ class TSPPreTrainedModel(PreTrainedModel):
|
|
32 |
module.bias.data.zero_()
|
33 |
module.weight.data.fill_(1.0)
|
34 |
|
|
|
35 |
# ====================================
|
36 |
# Pretraining Model
|
37 |
# ====================================
|
38 |
|
39 |
|
40 |
-
class
|
41 |
-
def __init__(self, config
|
42 |
super().__init__(config)
|
43 |
self.backbone = TSPModel(config)
|
44 |
-
if use_electra:
|
45 |
mlm_config = deepcopy(config)
|
46 |
-
mlm_config.hidden_size
|
47 |
-
mlm_config.intermediate_size
|
48 |
-
mlm_config.num_attention_heads
|
49 |
self.mlm_backbone = TSPModel(mlm_config)
|
50 |
self.mlm_head = MaskedLMHead(
|
51 |
mlm_config, word_embeddings=self.mlm_backbone.embeddings.word_embeddings
|
@@ -55,7 +56,10 @@ class TSPModelForPretraining(TSPPreTrainedModel):
|
|
55 |
self.rtd_head = ReplacedTokenDiscriminationHead(config)
|
56 |
else:
|
57 |
self.mlm_backbone = self.backbone
|
58 |
-
self.mlm_head = MaskedLMHead(
|
|
|
|
|
|
|
59 |
self.apply(self._init_weights)
|
60 |
|
61 |
def forward(self, *args, **kwargs):
|
@@ -63,40 +67,6 @@ class TSPModelForPretraining(TSPPreTrainedModel):
|
|
63 |
"Refer to the implementation of text structrue prediction task for how to use the model."
|
64 |
)
|
65 |
|
66 |
-
def mlm_forward(
|
67 |
-
self,
|
68 |
-
corrupted_ids, # <int>(B,L), partially masked/replaced input token ids
|
69 |
-
attention_mask, # <int>(B,L), 1 / 0 for tokens that are not attended/ attended
|
70 |
-
token_type_ids, # <int>(B,L), 0 / 1 corresponds to a segment A / B token
|
71 |
-
mlm_selected=None, # <bool>(B,L), True at mlm selected positiosns. Calculate logits at mlm selected positions if not None.
|
72 |
-
):
|
73 |
-
hidden_states = self.mlm_backbone(
|
74 |
-
input_ids=corrupted_ids,
|
75 |
-
attention_mask=attention_mask,
|
76 |
-
token_type_ids=token_type_ids,
|
77 |
-
) # (B,L,D)
|
78 |
-
return self.mlm_head(
|
79 |
-
hidden_states, is_selected=mlm_selected
|
80 |
-
) # (#mlm selected, vocab size)/ (B,L,vocab size)
|
81 |
-
|
82 |
-
def rtd_forward(
|
83 |
-
self,
|
84 |
-
corrupted_ids, # <int>(B,L), partially replaced input token ids
|
85 |
-
attention_mask, # <int>(B,L), 1 / 0 for tokens that are not attended/ attended
|
86 |
-
token_type_ids, # <int>(B,L), 0 / 1 corresponds to a segment A / B token
|
87 |
-
):
|
88 |
-
hidden_states = self.rtd_backbone(
|
89 |
-
input_ids=corrupted_ids,
|
90 |
-
attention_mask=attention_mask,
|
91 |
-
token_type_ids=token_type_ids,
|
92 |
-
) # (B,L,D)
|
93 |
-
return self.rtd_backbone(hidden_states) # (B,L)
|
94 |
-
|
95 |
-
def tsp_forward(
|
96 |
-
self, hidden_states, # (B,L,D)
|
97 |
-
):
|
98 |
-
raise NotImplementedError
|
99 |
-
|
100 |
|
101 |
class MaskedLMHead(nn.Module):
|
102 |
def __init__(self, config, word_embeddings=None):
|
@@ -135,6 +105,22 @@ class ReplacedTokenDiscriminationHead(nn.Module):
|
|
135 |
return x.squeeze(-1) # (B,L)
|
136 |
|
137 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
138 |
# ====================================
|
139 |
# Finetuning Model
|
140 |
# ====================================
|
@@ -164,8 +150,8 @@ class TSPModelForTokenClassification(TSPPreTrainedModel):
|
|
164 |
class TokenClassificationHead(nn.Module):
|
165 |
def __init__(self, config, num_classes):
|
166 |
super().__init__()
|
167 |
-
self.dropout = nn.Dropout(
|
168 |
-
self.classifier = nn.Linear(
|
169 |
|
170 |
def forward(self, x): # (B,L,D)
|
171 |
x = self.dropout(x) # (B,L,D)
|
@@ -213,6 +199,7 @@ class TSPModelForQuestionAnswering(TSPPreTrainedModel):
|
|
213 |
super().__init__()
|
214 |
self.backbone = TSPModel(config)
|
215 |
self.head = SequenceClassififcationHead(config, num_classes)
|
|
|
216 |
|
217 |
def forward(
|
218 |
self,
|
@@ -345,9 +332,6 @@ class SquadHead(nn.Module):
|
|
345 |
|
346 |
|
347 |
class TSPModel(TSPPreTrainedModel):
|
348 |
-
config_class = TSPConfig
|
349 |
-
base_model_prefix = "tsp"
|
350 |
-
|
351 |
def __init__(self, config):
|
352 |
super().__init__(config)
|
353 |
self.embeddings = Embeddings(config)
|
@@ -405,9 +389,9 @@ class Embeddings(nn.Module):
|
|
405 |
):
|
406 |
B, L = input_ids.shape
|
407 |
embeddings = self.word_embeddings(input_ids) # (B,L,E)
|
|
|
408 |
if hasattr(self, "position_embeddings"):
|
409 |
embeddings += self.position_embeddings.weight[None, :L, :]
|
410 |
-
embeddings += self.token_type_embeddings(token_type_ids)
|
411 |
embeddings = self.norm(embeddings) # (B,L,E)
|
412 |
embeddings = self.dropout(embeddings) # (B,L,E)
|
413 |
return embeddings # (B,L,E)
|
@@ -453,6 +437,8 @@ class MultiHeadSelfAttention(nn.Module):
|
|
453 |
self.o_proj = nn.Linear(config.hidden_size, config.hidden_size)
|
454 |
self.H = config.num_attention_heads
|
455 |
self.d = config.hidden_size // self.H
|
|
|
|
|
456 |
|
457 |
def forward(
|
458 |
self,
|
@@ -463,6 +449,8 @@ class MultiHeadSelfAttention(nn.Module):
|
|
463 |
query, key, value = (
|
464 |
self.mix_proj(x).view(B, L, H, 3 * d).transpose(1, 2).split(d, dim=-1)
|
465 |
) # (B,H,L,d),(B,H,L,d),(B,H,L,d)
|
|
|
|
|
466 |
output = self.attention(query, key, value, attention_mask) # (B,H,L,d)
|
467 |
output = self.o_proj(output.transpose(1, 2).reshape(B, L, D)) # (B,L,D)
|
468 |
return output # (B,L,D)
|
@@ -503,4 +491,43 @@ class FeedForwardNetwork(nn.Module):
|
|
503 |
return x # (B,L,D)
|
504 |
|
505 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
506 |
|
|
|
|
|
|
|
|
|
|
9 |
from torch import nn
|
10 |
import torch.nn.functional as F
|
11 |
from transformers import PreTrainedModel
|
12 |
+
from configuration_tsp import TSPConfig
|
13 |
|
14 |
|
15 |
class TSPPreTrainedModel(PreTrainedModel):
|
16 |
config_class = TSPConfig
|
17 |
+
base_model_prefix = "backbone"
|
18 |
|
19 |
def _init_weights(self, module):
|
20 |
"""Initialize the weights"""
|
|
|
32 |
module.bias.data.zero_()
|
33 |
module.weight.data.fill_(1.0)
|
34 |
|
35 |
+
|
36 |
# ====================================
|
37 |
# Pretraining Model
|
38 |
# ====================================
|
39 |
|
40 |
|
41 |
+
class TSPModelForPreTraining(TSPPreTrainedModel):
|
42 |
+
def __init__(self, config):
|
43 |
super().__init__(config)
|
44 |
self.backbone = TSPModel(config)
|
45 |
+
if config.use_electra:
|
46 |
mlm_config = deepcopy(config)
|
47 |
+
mlm_config.hidden_size //= config.electra_generator_size_divisor
|
48 |
+
mlm_config.intermediate_size //= config.electra_generator_size_divisor
|
49 |
+
mlm_config.num_attention_heads //= config.electra_generator_size_divisor
|
50 |
self.mlm_backbone = TSPModel(mlm_config)
|
51 |
self.mlm_head = MaskedLMHead(
|
52 |
mlm_config, word_embeddings=self.mlm_backbone.embeddings.word_embeddings
|
|
|
56 |
self.rtd_head = ReplacedTokenDiscriminationHead(config)
|
57 |
else:
|
58 |
self.mlm_backbone = self.backbone
|
59 |
+
self.mlm_head = MaskedLMHead(
|
60 |
+
config, word_embeddings=self.mlm_backbone.embeddings.word_embeddings
|
61 |
+
)
|
62 |
+
self.tsp_head = TextStructurePredictionHead(config)
|
63 |
self.apply(self._init_weights)
|
64 |
|
65 |
def forward(self, *args, **kwargs):
|
|
|
67 |
"Refer to the implementation of text structrue prediction task for how to use the model."
|
68 |
)
|
69 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
70 |
|
71 |
class MaskedLMHead(nn.Module):
|
72 |
def __init__(self, config, word_embeddings=None):
|
|
|
105 |
return x.squeeze(-1) # (B,L)
|
106 |
|
107 |
|
108 |
+
class TextStructurePredictionHead(nn.Module):
|
109 |
+
def __init__(self, config):
|
110 |
+
super().__init__()
|
111 |
+
self.linear1 = nn.Linear(config.hidden_size * 2, config.hidden_size * 2)
|
112 |
+
self.norm = nn.LayerNorm(config.hidden_size * 2)
|
113 |
+
self.linear2 = nn.Linear(config.hidden_size * 2, 6)
|
114 |
+
|
115 |
+
def forward(
|
116 |
+
self, x, # (...,2D)
|
117 |
+
):
|
118 |
+
x = self.linear1(x) # (...,2D)
|
119 |
+
x = F.gelu(x) # (...,2D)
|
120 |
+
x = self.norm(x) # (...,2D)
|
121 |
+
return self.linear2(x) # (...,C)
|
122 |
+
|
123 |
+
|
124 |
# ====================================
|
125 |
# Finetuning Model
|
126 |
# ====================================
|
|
|
150 |
class TokenClassificationHead(nn.Module):
|
151 |
def __init__(self, config, num_classes):
|
152 |
super().__init__()
|
153 |
+
self.dropout = nn.Dropout(config.dropout_prob)
|
154 |
+
self.classifier = nn.Linear(config.hidden_size, num_classes)
|
155 |
|
156 |
def forward(self, x): # (B,L,D)
|
157 |
x = self.dropout(x) # (B,L,D)
|
|
|
199 |
super().__init__()
|
200 |
self.backbone = TSPModel(config)
|
201 |
self.head = SequenceClassififcationHead(config, num_classes)
|
202 |
+
self.apply(self._init_weights)
|
203 |
|
204 |
def forward(
|
205 |
self,
|
|
|
332 |
|
333 |
|
334 |
class TSPModel(TSPPreTrainedModel):
|
|
|
|
|
|
|
335 |
def __init__(self, config):
|
336 |
super().__init__(config)
|
337 |
self.embeddings = Embeddings(config)
|
|
|
389 |
):
|
390 |
B, L = input_ids.shape
|
391 |
embeddings = self.word_embeddings(input_ids) # (B,L,E)
|
392 |
+
embeddings += self.token_type_embeddings(token_type_ids)
|
393 |
if hasattr(self, "position_embeddings"):
|
394 |
embeddings += self.position_embeddings.weight[None, :L, :]
|
|
|
395 |
embeddings = self.norm(embeddings) # (B,L,E)
|
396 |
embeddings = self.dropout(embeddings) # (B,L,E)
|
397 |
return embeddings # (B,L,E)
|
|
|
437 |
self.o_proj = nn.Linear(config.hidden_size, config.hidden_size)
|
438 |
self.H = config.num_attention_heads
|
439 |
self.d = config.hidden_size // self.H
|
440 |
+
if config.position_embedding_type == "rotary":
|
441 |
+
self.rotray_position_embeds = RotaryEmbedding(self.d)
|
442 |
|
443 |
def forward(
|
444 |
self,
|
|
|
449 |
query, key, value = (
|
450 |
self.mix_proj(x).view(B, L, H, 3 * d).transpose(1, 2).split(d, dim=-1)
|
451 |
) # (B,H,L,d),(B,H,L,d),(B,H,L,d)
|
452 |
+
if hasattr(self, "rotray_position_embeds"):
|
453 |
+
query, key = self.rotray_position_embeds(query, key)
|
454 |
output = self.attention(query, key, value, attention_mask) # (B,H,L,d)
|
455 |
output = self.o_proj(output.transpose(1, 2).reshape(B, L, D)) # (B,L,D)
|
456 |
return output # (B,L,D)
|
|
|
491 |
return x # (B,L,D)
|
492 |
|
493 |
|
494 |
+
class RotaryEmbedding(nn.Module):
|
495 |
+
seq_len_cached = 0
|
496 |
+
cos_cached = None
|
497 |
+
sin_cached = None
|
498 |
+
|
499 |
+
def __init__(self, dim):
|
500 |
+
super().__init__()
|
501 |
+
inv_freq = 1.0 / (10000 ** (torch.arange(0, dim, 2).float() / dim))
|
502 |
+
self.register_buffer("inv_freq", inv_freq)
|
503 |
+
|
504 |
+
def _forward(self, x): # (B,H,L,d)
|
505 |
+
# Get rotary embeddings on the fly
|
506 |
+
## create
|
507 |
+
seq_len = x.shape[2]
|
508 |
+
if seq_len > RotaryEmbedding.seq_len_cached:
|
509 |
+
RotaryEmbedding.seq_len_cached = seq_len
|
510 |
+
t = torch.arange(seq_len, device=x.device, dtype=self.inv_freq.dtype)
|
511 |
+
freqs = t.view(-1, 1) @ self.inv_freq.view(1, -1)
|
512 |
+
emb = torch.cat((freqs, freqs), dim=-1).to(x.device) # (L,d)
|
513 |
+
RotaryEmbedding.cos_cached = emb.cos()[None, None, :, :]
|
514 |
+
RotaryEmbedding.sin_cached = emb.sin()[None, None, :, :]
|
515 |
+
## take
|
516 |
+
if seq_len == RotaryEmbedding.seq_len_cached:
|
517 |
+
cos, sin = RotaryEmbedding.cos_cached, RotaryEmbedding.sin_cached
|
518 |
+
else:
|
519 |
+
cos, sin = (
|
520 |
+
RotaryEmbedding.cos_cached[:, :, :seq_len, :], # (1,1,L,d)
|
521 |
+
RotaryEmbedding.sin_cached[:, :, :seq_len, :], # (1,1,L,d)
|
522 |
+
)
|
523 |
+
|
524 |
+
# Apply rotary embeddings
|
525 |
+
sections = [x.shape[-1] // 2, x.shape[-1] - x.shape[-1] // 2]
|
526 |
+
x1, x2 = x.split(sections, dim=-1)
|
527 |
+
half_rotated_x = torch.cat((-x2, x1), dim=-1)
|
528 |
+
return (x * cos) + (half_rotated_x * sin)
|
529 |
|
530 |
+
def forward(
|
531 |
+
self, query, key, # (B,H,L,d) # (B,H,L,d)
|
532 |
+
):
|
533 |
+
return self._forward(query), self._forward(key)
|
pytorch_model.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8c8decb4de84befc5103d4b4b7c9ed0d61fc598ad859c30163e92107f76ea731
|
3 |
+
size 57777425
|