Spaces:
Paused
Paused
File size: 15,544 Bytes
ee6e328 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 |
# coding=utf-8
# Copyright 2020 The Google AI Language Team Authors, The HuggingFace Inc. team and Microsoft Corporation.
# Copyright (c) 2018, NVIDIA CORPORATION. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""PyTorch BERT model with Patience-based Early Exit. """
import logging
import torch
from torch import nn
from torch.nn import CrossEntropyLoss, MSELoss
from transformers.file_utils import add_start_docstrings, add_start_docstrings_to_model_forward
from transformers.models.bert.modeling_bert import (
BERT_INPUTS_DOCSTRING,
BERT_START_DOCSTRING,
BertEncoder,
BertModel,
BertPreTrainedModel,
)
logger = logging.getLogger(__name__)
class BertEncoderWithPabee(BertEncoder):
def adaptive_forward(self, hidden_states, current_layer, attention_mask=None, head_mask=None):
layer_outputs = self.layer[current_layer](hidden_states, attention_mask, head_mask[current_layer])
hidden_states = layer_outputs[0]
return hidden_states
@add_start_docstrings(
"The bare Bert Model transformer with PABEE outputting raw hidden-states without any specific head on top.",
BERT_START_DOCSTRING,
)
class BertModelWithPabee(BertModel):
"""
The model can behave as an encoder (with only self-attention) as well
as a decoder, in which case a layer of cross-attention is added between
the self-attention layers, following the architecture described in `Attention is all you need`_ by Ashish Vaswani,
Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser and Illia Polosukhin.
To behave as a decoder the model needs to be initialized with the
:obj:`is_decoder` argument of the configuration set to :obj:`True`; an
:obj:`encoder_hidden_states` is expected as an input to the forward pass.
.. _`Attention is all you need`:
https://arxiv.org/abs/1706.03762
"""
def __init__(self, config):
super().__init__(config)
self.encoder = BertEncoderWithPabee(config)
self.init_weights()
self.patience = 0
self.inference_instances_num = 0
self.inference_layers_num = 0
self.regression_threshold = 0
def set_regression_threshold(self, threshold):
self.regression_threshold = threshold
def set_patience(self, patience):
self.patience = patience
def reset_stats(self):
self.inference_instances_num = 0
self.inference_layers_num = 0
def log_stats(self):
avg_inf_layers = self.inference_layers_num / self.inference_instances_num
message = (
f"*** Patience = {self.patience} Avg. Inference Layers = {avg_inf_layers:.2f} Speed Up ="
f" {1 - avg_inf_layers / self.config.num_hidden_layers:.2f} ***"
)
print(message)
@add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING)
def forward(
self,
input_ids=None,
attention_mask=None,
token_type_ids=None,
position_ids=None,
head_mask=None,
inputs_embeds=None,
encoder_hidden_states=None,
encoder_attention_mask=None,
output_dropout=None,
output_layers=None,
regression=False,
):
r"""
Return:
:obj:`tuple(torch.FloatTensor)` comprising various elements depending on the configuration (:class:`~transformers.BertConfig`) and inputs:
last_hidden_state (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length, hidden_size)`):
Sequence of hidden-states at the output of the last layer of the model.
pooler_output (:obj:`torch.FloatTensor`: of shape :obj:`(batch_size, hidden_size)`):
Last layer hidden-state of the first token of the sequence (classification token)
further processed by a Linear layer and a Tanh activation function. The Linear
layer weights are trained from the next sentence prediction (classification)
objective during pre-training.
This output is usually *not* a good summary
of the semantic content of the input, you're often better with averaging or pooling
the sequence of hidden-states for the whole input sequence.
hidden_states (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``config.output_hidden_states=True``):
Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer)
of shape :obj:`(batch_size, sequence_length, hidden_size)`.
Hidden-states of the model at the output of each layer plus the initial embedding outputs.
attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
heads.
"""
if input_ids is not None and inputs_embeds is not None:
raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
elif input_ids is not None:
input_shape = input_ids.size()
elif inputs_embeds is not None:
input_shape = inputs_embeds.size()[:-1]
else:
raise ValueError("You have to specify either input_ids or inputs_embeds")
device = input_ids.device if input_ids is not None else inputs_embeds.device
if attention_mask is None:
attention_mask = torch.ones(input_shape, device=device)
if token_type_ids is None:
token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=device)
# We can provide a self-attention mask of dimensions [batch_size, from_seq_length, to_seq_length]
# ourselves in which case we just need to make it broadcastable to all heads.
extended_attention_mask: torch.Tensor = self.get_extended_attention_mask(attention_mask, input_shape, device)
# If a 2D ou 3D attention mask is provided for the cross-attention
# we need to make broadcastable to [batch_size, num_heads, seq_length, seq_length]
if self.config.is_decoder and encoder_hidden_states is not None:
encoder_batch_size, encoder_sequence_length, _ = encoder_hidden_states.size()
encoder_hidden_shape = (encoder_batch_size, encoder_sequence_length)
if encoder_attention_mask is None:
encoder_attention_mask = torch.ones(encoder_hidden_shape, device=device)
encoder_extended_attention_mask = self.invert_attention_mask(encoder_attention_mask)
else:
encoder_extended_attention_mask = None
# Prepare head mask if needed
# 1.0 in head_mask indicate we keep the head
# attention_probs has shape bsz x n_heads x N x N
# input head_mask has shape [num_heads] or [num_hidden_layers x num_heads]
# and head_mask is converted to shape [num_hidden_layers x batch x num_heads x seq_length x seq_length]
head_mask = self.get_head_mask(head_mask, self.config.num_hidden_layers)
embedding_output = self.embeddings(
input_ids=input_ids, position_ids=position_ids, token_type_ids=token_type_ids, inputs_embeds=inputs_embeds
)
encoder_outputs = embedding_output
if self.training:
res = []
for i in range(self.config.num_hidden_layers):
encoder_outputs = self.encoder.adaptive_forward(
encoder_outputs, current_layer=i, attention_mask=extended_attention_mask, head_mask=head_mask
)
pooled_output = self.pooler(encoder_outputs)
logits = output_layers[i](output_dropout(pooled_output))
res.append(logits)
elif self.patience == 0: # Use all layers for inference
encoder_outputs = self.encoder(
embedding_output,
attention_mask=extended_attention_mask,
head_mask=head_mask,
encoder_hidden_states=encoder_hidden_states,
encoder_attention_mask=encoder_extended_attention_mask,
)
pooled_output = self.pooler(encoder_outputs[0])
res = [output_layers[self.config.num_hidden_layers - 1](pooled_output)]
else:
patient_counter = 0
patient_result = None
calculated_layer_num = 0
for i in range(self.config.num_hidden_layers):
calculated_layer_num += 1
encoder_outputs = self.encoder.adaptive_forward(
encoder_outputs, current_layer=i, attention_mask=extended_attention_mask, head_mask=head_mask
)
pooled_output = self.pooler(encoder_outputs)
logits = output_layers[i](pooled_output)
if regression:
labels = logits.detach()
if patient_result is not None:
patient_labels = patient_result.detach()
if (patient_result is not None) and torch.abs(patient_result - labels) < self.regression_threshold:
patient_counter += 1
else:
patient_counter = 0
else:
labels = logits.detach().argmax(dim=1)
if patient_result is not None:
patient_labels = patient_result.detach().argmax(dim=1)
if (patient_result is not None) and torch.all(labels.eq(patient_labels)):
patient_counter += 1
else:
patient_counter = 0
patient_result = logits
if patient_counter == self.patience:
break
res = [patient_result]
self.inference_layers_num += calculated_layer_num
self.inference_instances_num += 1
return res
@add_start_docstrings(
"""Bert Model transformer with PABEE and a sequence classification/regression head on top (a linear layer on top of
the pooled output) e.g. for GLUE tasks. """,
BERT_START_DOCSTRING,
)
class BertForSequenceClassificationWithPabee(BertPreTrainedModel):
def __init__(self, config):
super().__init__(config)
self.num_labels = config.num_labels
self.bert = BertModelWithPabee(config)
self.dropout = nn.Dropout(config.hidden_dropout_prob)
self.classifiers = nn.ModuleList(
[nn.Linear(config.hidden_size, self.config.num_labels) for _ in range(config.num_hidden_layers)]
)
self.init_weights()
@add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING)
def forward(
self,
input_ids=None,
attention_mask=None,
token_type_ids=None,
position_ids=None,
head_mask=None,
inputs_embeds=None,
labels=None,
):
r"""
labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size,)`, `optional`):
Labels for computing the sequence classification/regression loss.
Indices should be in :obj:`[0, ..., config.num_labels - 1]`.
If :obj:`config.num_labels == 1` a regression loss is computed (Mean-Square loss),
If :obj:`config.num_labels > 1` a classification loss is computed (Cross-Entropy).
Returns:
:obj:`tuple(torch.FloatTensor)` comprising various elements depending on the configuration (:class:`~transformers.BertConfig`) and inputs:
loss (:obj:`torch.FloatTensor` of shape :obj:`(1,)`, `optional`, returned when :obj:`label` is provided):
Classification (or regression if config.num_labels==1) loss.
logits (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, config.num_labels)`):
Classification (or regression if config.num_labels==1) scores (before SoftMax).
hidden_states (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``config.output_hidden_states=True``):
Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer)
of shape :obj:`(batch_size, sequence_length, hidden_size)`.
Hidden-states of the model at the output of each layer plus the initial embedding outputs.
attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
heads.
Examples::
from transformers import BertTokenizer, BertForSequenceClassification
from pabee import BertForSequenceClassificationWithPabee
from torch import nn
import torch
tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
model = BertForSequenceClassificationWithPabee.from_pretrained('bert-base-uncased')
input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute", add_special_tokens=True)).unsqueeze(0) # Batch size 1
labels = torch.tensor([1]).unsqueeze(0) # Batch size 1
outputs = model(input_ids, labels=labels)
loss, logits = outputs[:2]
"""
logits = self.bert(
input_ids=input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
position_ids=position_ids,
head_mask=head_mask,
inputs_embeds=inputs_embeds,
output_dropout=self.dropout,
output_layers=self.classifiers,
regression=self.num_labels == 1,
)
outputs = (logits[-1],)
if labels is not None:
total_loss = None
total_weights = 0
for ix, logits_item in enumerate(logits):
if self.num_labels == 1:
# We are doing regression
loss_fct = MSELoss()
loss = loss_fct(logits_item.view(-1), labels.view(-1))
else:
loss_fct = CrossEntropyLoss()
loss = loss_fct(logits_item.view(-1, self.num_labels), labels.view(-1))
if total_loss is None:
total_loss = loss
else:
total_loss += loss * (ix + 1)
total_weights += ix + 1
outputs = (total_loss / total_weights,) + outputs
return outputs
|