将 LIME 解释应用于我的微调 BERT 以进行序列分类模型？
Posted
技术标签:
【中文标题】将 LIME 解释应用于我的微调 BERT 以进行序列分类模型？【英文标题】：Applying LIME interpretation on my fine-tuned BERT for sequence classification model? 【发布时间】：2020-10-22 14:45:03 【问题描述】：我针对特定任务的序列分类对 BERT 进行了微调，我想应用 LIME 解释来查看每个标记如何有助于分类到特定标签，因为 LIME 将分类器作为黑盒处理。我从可用的在线代码中制作了一个组合代码，如下所示：
# coding=utf-8
# Copyright 2018 The Google AI Language Team Authors and The HugginFace Inc. team.
# 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.
#This code is a modification of the original run_examples.py script from the repository 'https://#github.com/wanghm92/pytorch-pretrained-BERT'. Modified for research purposes by Andraž P.
import logging
import csv
import argparse
import numpy as np
import torch
from torch.utils.data import TensorDataset, DataLoader, RandomSampler, SequentialSampler
from torch.utils.data.distributed import DistributedSampler
from pytorch_pretrained_bert.modeling import BertForSequenceClassification
from pytorch_pretrained_bert.tokenization import BertTokenizer
from tqdm import tqdm
logging.basicConfig(format = '%(asctime)s - %(levelname)s - %(name)s - %(message)s',
datefmt = '%m/%d/%Y %H:%M:%S',
level = logging.INFO)
logger = logging.getLogger(__name__)
class InputExample(object):
"""A single training/test example for simple sequence classification."""
def __init__(self, guid, text_a, text_b=None, label=None):
"""Constructs a InputExample.
Args:
guid: Unique id for the example.
text_a: string. The untokenized text of the first sequence. For single
sequence tasks, only this sequence must be specified.
text_b: (Optional) string. The untokenized text of the second sequence.
Only must be specified for sequence pair tasks.
label: (Optional) string. The label of the example. This should be
specified for train and dev examples, but not for test examples.
"""
self.guid = guid
self.text_a = text_a
self.text_b = text_b
self.label = label
class InputFeatures(object):
"""A single set of features of data."""
def __init__(self, input_ids, input_mask, segment_ids, label_id):
self.input_ids = input_ids
self.input_mask = input_mask
self.segment_ids = segment_ids
self.label_id = label_id
class DataProcessor(object):
"""Base class for data converters for sequence classification data sets."""
def get_train_examples(self, data_dir):
"""Gets a collection of `InputExample`s for the train set."""
raise NotImplementedError()
def get_dev_examples(self, data_dir):
"""Gets a collection of `InputExample`s for the dev set."""
raise NotImplementedError()
def get_labels(self):
"""Gets the list of labels for this data set."""
raise NotImplementedError()
@classmethod
def _read_tsv(cls, input_file, quotechar=None):
"""Reads a tab separated value file."""
with open(input_file, "r", encoding='utf-8') as f:
reader = csv.reader(f, delimiter="\t", quotechar=quotechar)
lines = []
for line in reader:
lines.append(line)
return lines
class SemEvalProcessor(DataProcessor):
def get_train_examples(self, data_dir):
"""See base class."""
logger.info("LOOKING AT ".format(data_dir))
return self._create_examples(
self._read_tsv(data_dir), "train")
def get_dev_examples(self, data_dir):
"""See base class."""
return self._create_examples(
self._read_tsv(data_dir), "dev")
def get_test_examples(self, data_dir):
logger.info("LOOKING AT ".format(data_dir))
return self._create_examples(
self._read_tsv(data_dir), "test")
def get_labels(self):
"""See base class."""
return ["0", "1"]
def create_examples(self, line, set_type='test'):
examples = []
if set_type == 'test':
guid = "%s" % (set_type)
text_a = line
text_b = None
label = "0"
examples.append(InputExample(guid=guid, text_a=text_a, text_b=text_b, label=label))
return examples[0]
class Predictions:
def convert_examples_to_features(self,example, label_list, max_seq_length, tokenizer):
"""Loads a data file into a list of `InputBatch`s."""
if label_list:
label_map = label : i for i, label in enumerate(label_list)
else:
label_map = "0": i for i in range(1)
features = []
tokens_a = tokenizer.tokenize(example.text_a)
tokens_b = None
if example.text_b:
tokens_b = tokenizer.tokenize(example.text_b)
# Modifies `tokens_a` and `tokens_b` in place so that the total
# length is less than the specified length.
# Account for [CLS], [SEP], [SEP] with "- 3"
_truncate_seq_pair(tokens_a, tokens_b, max_seq_length - 3)
else:
# Account for [CLS] and [SEP] with "- 2"
if len(tokens_a) > max_seq_length - 2:
tokens_a = tokens_a[:(max_seq_length - 2)]
# The convention in BERT is:
# (a) For sequence pairs:
# tokens: [CLS] is this jack ##son ##ville ? [SEP] no it is not . [SEP]
# type_ids: 0 0 0 0 0 0 0 0 1 1 1 1 1 1
# (b) For single sequences:
# tokens: [CLS] the dog is hairy . [SEP]
# type_ids: 0 0 0 0 0 0 0
#
# Where "type_ids" are used to indicate whether this is the first
# sequence or the second sequence. The embedding vectors for `type=0` and
# `type=1` were learned during pre-training and are added to the wordpiece
# embedding vector (and position vector). This is not *strictly* necessary
# since the [SEP] token unambigiously separates the sequences, but it makes
# it easier for the model to learn the concept of sequences.
#
# For classification tasks, the first vector (corresponding to [CLS]) is
# used as as the "sentence vector". Note that this only makes sense because
# the entire model is fine-tuned.
tokens = ["[CLS]"] + tokens_a + ["[SEP]"]
segment_ids = [0] * len(tokens)
if tokens_b:
tokens += tokens_b + ["[SEP]"]
segment_ids += [1] * (len(tokens_b) + 1)
input_ids = tokenizer.convert_tokens_to_ids(tokens)
# The mask has 1 for real tokens and 0 for padding tokens. Only real
# tokens are attended to.
input_mask = [1] * len(input_ids)
# Zero-pad up to the sequence length.
padding = [0] * (max_seq_length - len(input_ids))
input_ids += padding
input_mask += padding
segment_ids += padding
assert len(input_ids) == max_seq_length
assert len(input_mask) == max_seq_length
assert len(segment_ids) == max_seq_length
label_id = label_map[example.label]
logger.info("*** Example ***")
logger.info("guid: %s" % (example.guid))
logger.info("tokens: %s" % " ".join(
[str(x) for x in tokens]))
logger.info("input_ids: %s" % " ".join([str(x) for x in input_ids]))
logger.info("input_mask: %s" % " ".join([str(x) for x in input_mask]))
logger.info(
"segment_ids: %s" % " ".join([str(x) for x in segment_ids]))
logger.info("label: %s (id = %d)" % (example.label, label_id))
features.append(
InputFeatures(input_ids=input_ids,
input_mask=input_mask,
segment_ids=segment_ids,
label_id=label_id))
return features
def _truncate_seq_pair(tokens_a, tokens_b, max_length):
"""Truncates a sequence pair in place to the maximum length."""
# This is a simple heuristic which will always truncate the longer sequence
# one token at a time. This makes more sense than truncating an equal percent
# of tokens from each, since if one sequence is very short then each token
# that's truncated likely contains more information than a longer sequence.
while True:
total_length = len(tokens_a) + len(tokens_b)
if total_length <= max_length:
break
if len(tokens_a) > len(tokens_b):
tokens_a.pop()
else:
tokens_b.pop()
#def accuracy(out, labels):
# outputs = np.argmax(out, axis=1)
# return np.sum(outputs == labels)
#def warmup_linear(x, warmup=0.002):
# if x < warmup:
# return x/warmup
# return 1.0 - x
def predict(self, text):
examples = []
print(text)
for example in text:
test_example = processor.create_examples(example)
test_features = self.convert_examples_to_features(test_example, None, args.max_seq_length, tokenizer)
examples.append(test_features)
logger.info("***** Running prediction *****")
#logger.info(" test_example = %", test_example)
logger.info(" Batch size = %d", args.predict_batch_size)
all_input_ids = torch.tensor([f.input_ids for f in test_features], dtype=torch.long)
all_input_mask = torch.tensor([f.input_mask for f in test_features], dtype=torch.long)
all_segment_ids = torch.tensor([f.segment_ids for f in test_features], dtype=torch.long)
all_label_ids = torch.tensor([f.label_id for f in test_features], dtype=torch.long)
test_data = TensorDataset(all_input_ids, all_input_mask, all_segment_ids, all_label_ids)
# Run prediction for full data
test_sampler = SequentialSampler(test_data)
test_dataloader = DataLoader(test_data, sampler=test_sampler, batch_size=args.predict_batch_size)
model.eval()
all_preds = []
all_ids = []
results=[]
for input_ids, input_mask, segment_ids, label_ids in tqdm(test_dataloader, desc="Predicting"):
input_ids = input_ids.to(device)
input_mask = input_mask.to(device)
segment_ids = segment_ids.to(device)
with torch.no_grad():
outputs = model(input_ids, segment_ids, input_mask)
logits = outputs[0]
print("logits,logits.shape",logits)
#preds = np.argmax(logits).tolist()
logits = F.softmax(logits, dim = 1)
results.append(logits.cpu().detach().numpy()[0])
#logits = logits.detach().cpu().numpy()[0]
input_ids = input_ids.to('cpu').numpy().tolist()
#all_preds.extend(preds)
all_ids.extend(input_ids)
results_array = np.array(results)
logger.info("results_array",results_array.shape)
return np.array(logits)
if __name__ == "__main__":
import pandas as pd
import numpy as np
import torch
import torch.nn.functional as F
from lime.lime_text import LimeTextExplainer
import logging
parser = argparse.ArgumentParser()
## Required parameters
parser.add_argument("--data_path",
type=str,
help="The path of the input data file. The data should be in the .tsv format.")
parser.add_argument("--bert_model", default='bert-base-uncased', type=str,
help="Bert pre-trained model selected in the list: bert-base-uncased, "
"bert-large-uncased, bert-base-cased, bert-large-cased, bert-base-multilingual-uncased, "
"bert-base-multilingual-cased, bert-base-chinese.")
parser.add_argument("--task_name",
default='SEMEVAL',
type=str,
help="The name of the task.")
parser.add_argument("--output_dir",
default='results/bert_output.tsv',
type=str,
help="The output results file.")
parser.add_argument("--model_path",
default='results/semeval_output/pytorch_model_task.bin',
type=str,
help="The path to the trained model file.")
## Other parameters
parser.add_argument("--max_seq_length",
default=128,
type=int,
help="The maximum total input sequence length after WordPiece tokenization. \n"
"Sequences longer than this will be truncated, and sequences shorter \n"
"than this will be padded.")
parser.add_argument("--do_lower_case",
action='store_true',
default=True,
help="Set this flag if you are using an uncased model.")
parser.add_argument("--predict_batch_size",
default=16,
type=int,
help="Total batch size for prediction.")
parser.add_argument("--no_cuda",
action='store_true',
help="Whether not to use CUDA when available")
parser.add_argument("--local_rank",
type=int,
default=-1,
help="local_rank for distributed training on gpus")
parser.add_argument('--fp16',
action='store_true',
help="Whether to use 16-bit float precision instead of 32-bit")
args = parser.parse_args()
processors =
"semeval": SemEvalProcessor,
num_labels_task =
"semeval": 2,
if args.local_rank == -1 or args.no_cuda:
device = torch.device("cuda" if torch.cuda.is_available() and not args.no_cuda else "cpu")
n_gpu = torch.cuda.device_count()
else:
torch.cuda.set_device(args.local_rank)
device = torch.device("cuda", args.local_rank)
n_gpu = 1
# Initializes the distributed backend which will take care of sychronizing nodes/GPUs
torch.distributed.init_process_group(backend='nccl')
logger.info("device: n_gpu: , distributed training: , 16-bits training: ".format(
device, n_gpu, bool(args.local_rank != -1), args.fp16))
task_name = args.task_name.lower()
if task_name not in processors:
raise ValueError("Task not found: %s" % (task_name))
processor = processors[task_name]()
num_labels = num_labels_task[task_name]
#label_list = processor.get_labels()
# loading a tokenizer
tokenizer = BertTokenizer.from_pretrained(args.bert_model, do_lower_case=args.do_lower_case)
# Lodaing a model
model_state_dict = torch.load(args.model_path)
model = BertForSequenceClassification.from_pretrained(args.bert_model, state_dict=model_state_dict,
num_labels=num_labels)
model.to(device)
predictions = Predictions()
#model_path = "models/mrpc"
#bert_model_class = "bert"
#prediction = Prediction(bert_model_class, model_path,
# lower_case = True, seq_length = 512)
label_names = [0, 1]
train_df = pd.read_csv(args.data_path, sep = '\t')
#test_examples = processor.get_test_examples(texts)
# guids = [example.guid for example in test_examples]
#resultpredict=predict(test_examples)
logger.info("Applying LIME")
explainer = LimeTextExplainer(class_names=label_names)
train_ls = train_df['text'].tolist()
train_ls=train_ls[0]
example=train_ls
print(type(predictions.predict))
for example in train_ls:
print('example train_ls',example)
exp = explainer.explain_instance(example, predictions.predict)
words = exp.as_list()
logger.info("Fininsh Applying LIME")
当我运行代码时，由于 predictions.predict 输入，它永远不会给我输出，总是有我无法计算的尺寸错误，我应该以最简单的形式输入 explainer.explain_instance 什么？ 我应该输入 ('text',[[0.9867 0.1243]]) 吗？ 或者 （'文本'，[[0.9867 0.1243] [0.7651 0.3459] [0.3254 0.775]]） 要不然是啥？ 我收到了这个错误
Traceback (most recent call last):
File "/content/drive/My Drive/BERT CODE/Try_LIME.py", line 442, in <module>
exp = explainer.explain_instance(example, predictions.predict)
File "/usr/local/lib/python3.6/dist-packages/lime/lime_text.py", line 415, in explain_instance
distance_metric=distance_metric)
File "/usr/local/lib/python3.6/dist-packages/lime/lime_text.py", line 482, in __data_labels_distances
labels = classifier_fn(inverse_data)
File "/content/drive/My Drive/BERT CODE/Try_LIME.py", line 273, in predict
logits = F.softmax(logits, dim = 1)
File "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py", line 1498, in softmax
ret = input.softmax(dim)
IndexError: Dimension out of range (expected to be in range of [-1, 0], but got 1)
提前致谢
【问题讨论】：
鉴于我是python和bert的初学者 我也收到了这个错误： 【参考方案1】：例子太复杂了。
您可以在检查您为什么不工作时遵循简单的方法（在 jupyter-lab 中）
import numpy as np
import lime
import torch
import torch.nn.functional as F
from lime.lime_text import LimeTextExplainer
from transformers import AutoTokenizer, AutoModelForSequenceClassification
filename_model = 'ProsusAI/finbert'
tokenizer = AutoTokenizer.from_pretrained(filename_model)
model = AutoModelForSequenceClassification.from_pretrained(filename_model)
class_names = ['positive','negative', 'neutral']
def predictor(texts):
outputs = model(**tokenizer(texts, return_tensors="pt", padding=True))
tensor_logits = outputs[0]
probas = F.softmax(tensor_logits).detach().numpy()
return probas
text = 'Building more bypasses will help the environment by reducing pollution and traffic jams in towns and cities.'
print(tokenizer(text, return_tensors='pt', padding=True))
explainer = LimeTextExplainer(class_names=class_names)
exp = explainer.explain_instance(text, predictor, num_features=20, num_samples=2000)
exp.show_in_notebook(text=text)
【讨论】：
以上是关于将 LIME 解释应用于我的微调 BERT 以进行序列分类模型？的主要内容，如果未能解决你的问题，请参考以下文章
输入文件应该如何格式化以进行语言模型微调（BERT 通过 Huggingface Transformers）？