feat: added further GLUE models
Browse files- modeling_for_glue.py +160 -1
modeling_for_glue.py
CHANGED
@@ -3,7 +3,7 @@ from typing import Optional, Union, Tuple
|
|
3 |
import torch
|
4 |
from torch import nn
|
5 |
from torch.nn import CrossEntropyLoss, MSELoss, BCEWithLogitsLoss
|
6 |
-
from transformers.modeling_outputs import SequenceClassifierOutput
|
7 |
|
8 |
from .modeling_bert import BertPreTrainedModel, BertModel
|
9 |
from .configuration_bert import JinaBertConfig
|
@@ -102,3 +102,162 @@ class BertForSequenceClassification(BertPreTrainedModel):
|
|
102 |
hidden_states=outputs.hidden_states,
|
103 |
attentions=outputs.attentions,
|
104 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
import torch
|
4 |
from torch import nn
|
5 |
from torch.nn import CrossEntropyLoss, MSELoss, BCEWithLogitsLoss
|
6 |
+
from transformers.modeling_outputs import SequenceClassifierOutput, QuestionAnsweringModelOutput, TokenClassifierOutput
|
7 |
|
8 |
from .modeling_bert import BertPreTrainedModel, BertModel
|
9 |
from .configuration_bert import JinaBertConfig
|
|
|
102 |
hidden_states=outputs.hidden_states,
|
103 |
attentions=outputs.attentions,
|
104 |
)
|
105 |
+
|
106 |
+
class BertForQuestionAnswering(BertPreTrainedModel):
|
107 |
+
def __init__(self, config: JinaBertConfig):
|
108 |
+
super().__init__(config)
|
109 |
+
self.num_labels = config.num_labels
|
110 |
+
|
111 |
+
self.bert = BertModel(config, add_pooling_layer=False)
|
112 |
+
self.qa_outputs = nn.Linear(config.hidden_size, config.num_labels)
|
113 |
+
|
114 |
+
# Initialize weights and apply final processing
|
115 |
+
self.post_init()
|
116 |
+
|
117 |
+
def forward(
|
118 |
+
self,
|
119 |
+
input_ids: Optional[torch.Tensor] = None,
|
120 |
+
attention_mask: Optional[torch.Tensor] = None,
|
121 |
+
token_type_ids: Optional[torch.Tensor] = None,
|
122 |
+
position_ids: Optional[torch.Tensor] = None,
|
123 |
+
head_mask: Optional[torch.Tensor] = None,
|
124 |
+
inputs_embeds: Optional[torch.Tensor] = None,
|
125 |
+
start_positions: Optional[torch.Tensor] = None,
|
126 |
+
end_positions: Optional[torch.Tensor] = None,
|
127 |
+
output_attentions: Optional[bool] = None,
|
128 |
+
output_hidden_states: Optional[bool] = None,
|
129 |
+
return_dict: Optional[bool] = None,
|
130 |
+
) -> Union[Tuple[torch.Tensor], QuestionAnsweringModelOutput]:
|
131 |
+
r"""
|
132 |
+
start_positions (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
|
133 |
+
Labels for position (index) of the start of the labelled span for computing the token classification loss.
|
134 |
+
Positions are clamped to the length of the sequence (`sequence_length`). Position outside of the sequence
|
135 |
+
are not taken into account for computing the loss.
|
136 |
+
end_positions (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
|
137 |
+
Labels for position (index) of the end of the labelled span for computing the token classification loss.
|
138 |
+
Positions are clamped to the length of the sequence (`sequence_length`). Position outside of the sequence
|
139 |
+
are not taken into account for computing the loss.
|
140 |
+
"""
|
141 |
+
return_dict = (
|
142 |
+
return_dict if return_dict is not None else self.config.use_return_dict
|
143 |
+
)
|
144 |
+
|
145 |
+
outputs = self.bert(
|
146 |
+
input_ids,
|
147 |
+
attention_mask=attention_mask,
|
148 |
+
token_type_ids=token_type_ids,
|
149 |
+
position_ids=position_ids,
|
150 |
+
head_mask=head_mask,
|
151 |
+
inputs_embeds=inputs_embeds,
|
152 |
+
output_attentions=output_attentions,
|
153 |
+
output_hidden_states=output_hidden_states,
|
154 |
+
return_dict=return_dict,
|
155 |
+
)
|
156 |
+
|
157 |
+
sequence_output = outputs[0]
|
158 |
+
|
159 |
+
logits = self.qa_outputs(sequence_output)
|
160 |
+
start_logits, end_logits = logits.split(1, dim=-1)
|
161 |
+
start_logits = start_logits.squeeze(-1).contiguous()
|
162 |
+
end_logits = end_logits.squeeze(-1).contiguous()
|
163 |
+
|
164 |
+
total_loss = None
|
165 |
+
if start_positions is not None and end_positions is not None:
|
166 |
+
# If we are on multi-GPU, split add a dimension
|
167 |
+
if len(start_positions.size()) > 1:
|
168 |
+
start_positions = start_positions.squeeze(-1)
|
169 |
+
if len(end_positions.size()) > 1:
|
170 |
+
end_positions = end_positions.squeeze(-1)
|
171 |
+
# sometimes the start/end positions are outside our model inputs, we ignore these terms
|
172 |
+
ignored_index = start_logits.size(1)
|
173 |
+
start_positions = start_positions.clamp(0, ignored_index)
|
174 |
+
end_positions = end_positions.clamp(0, ignored_index)
|
175 |
+
|
176 |
+
loss_fct = CrossEntropyLoss(ignore_index=ignored_index)
|
177 |
+
start_loss = loss_fct(start_logits, start_positions)
|
178 |
+
end_loss = loss_fct(end_logits, end_positions)
|
179 |
+
total_loss = (start_loss + end_loss) / 2
|
180 |
+
|
181 |
+
if not return_dict:
|
182 |
+
output = (start_logits, end_logits) + outputs[2:]
|
183 |
+
return ((total_loss,) + output) if total_loss is not None else output
|
184 |
+
|
185 |
+
return QuestionAnsweringModelOutput(
|
186 |
+
loss=total_loss,
|
187 |
+
start_logits=start_logits,
|
188 |
+
end_logits=end_logits,
|
189 |
+
hidden_states=outputs.hidden_states,
|
190 |
+
attentions=outputs.attentions,
|
191 |
+
)
|
192 |
+
|
193 |
+
|
194 |
+
class BertForTokenClassification(BertPreTrainedModel):
|
195 |
+
def __init__(self, config: JinaBertConfig):
|
196 |
+
super().__init__(config)
|
197 |
+
self.num_labels = config.num_labels
|
198 |
+
|
199 |
+
self.bert = BertModel(config, add_pooling_layer=False)
|
200 |
+
classifier_dropout = (
|
201 |
+
config.classifier_dropout
|
202 |
+
if config.classifier_dropout is not None
|
203 |
+
else config.hidden_dropout_prob
|
204 |
+
)
|
205 |
+
self.dropout = nn.Dropout(classifier_dropout)
|
206 |
+
self.classifier = nn.Linear(config.hidden_size, config.num_labels)
|
207 |
+
|
208 |
+
# Initialize weights and apply final processing
|
209 |
+
self.post_init()
|
210 |
+
|
211 |
+
def forward(
|
212 |
+
self,
|
213 |
+
input_ids: Optional[torch.Tensor] = None,
|
214 |
+
attention_mask: Optional[torch.Tensor] = None,
|
215 |
+
token_type_ids: Optional[torch.Tensor] = None,
|
216 |
+
position_ids: Optional[torch.Tensor] = None,
|
217 |
+
head_mask: Optional[torch.Tensor] = None,
|
218 |
+
inputs_embeds: Optional[torch.Tensor] = None,
|
219 |
+
labels: Optional[torch.Tensor] = None,
|
220 |
+
output_attentions: Optional[bool] = None,
|
221 |
+
output_hidden_states: Optional[bool] = None,
|
222 |
+
return_dict: Optional[bool] = None,
|
223 |
+
) -> Union[Tuple[torch.Tensor], TokenClassifierOutput]:
|
224 |
+
r"""
|
225 |
+
labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
|
226 |
+
Labels for computing the token classification loss. Indices should be in `[0, ..., config.num_labels - 1]`.
|
227 |
+
"""
|
228 |
+
return_dict = (
|
229 |
+
return_dict if return_dict is not None else self.config.use_return_dict
|
230 |
+
)
|
231 |
+
|
232 |
+
outputs = self.bert(
|
233 |
+
input_ids,
|
234 |
+
attention_mask=attention_mask,
|
235 |
+
token_type_ids=token_type_ids,
|
236 |
+
position_ids=position_ids,
|
237 |
+
head_mask=head_mask,
|
238 |
+
inputs_embeds=inputs_embeds,
|
239 |
+
output_attentions=output_attentions,
|
240 |
+
output_hidden_states=output_hidden_states,
|
241 |
+
return_dict=return_dict,
|
242 |
+
)
|
243 |
+
|
244 |
+
sequence_output = outputs[0]
|
245 |
+
|
246 |
+
sequence_output = self.dropout(sequence_output)
|
247 |
+
logits = self.classifier(sequence_output)
|
248 |
+
|
249 |
+
loss = None
|
250 |
+
if labels is not None:
|
251 |
+
loss_fct = CrossEntropyLoss()
|
252 |
+
loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
|
253 |
+
|
254 |
+
if not return_dict:
|
255 |
+
output = (logits,) + outputs[2:]
|
256 |
+
return ((loss,) + output) if loss is not None else output
|
257 |
+
|
258 |
+
return TokenClassifierOutput(
|
259 |
+
loss=loss,
|
260 |
+
logits=logits,
|
261 |
+
hidden_states=outputs.hidden_states,
|
262 |
+
attentions=outputs.attentions,
|
263 |
+
)
|