标签：bert 调用 keras self len x2 model X1 data
mark一下使用keras_bert调用bert的简单方法，来源于某位大佬，在此表示感谢（找不到来源之处了）。
import json
import numpy as np
import pandas as pd
from keras_bert import load_trained_model_from_checkpoint, Tokenizer, load_vocabulary # 超参数
from keras.layers import *
from keras.models import Model
import keras.backend as K
from keras.optimizers import Adam
maxlen = 100
batch_size = 8
droup_out_rate = 0.5
learning_rate = 1e-5
epochs = 1
#
path_prefix = "E:/NlpExternalPkg"
# # 预训练模型目录
config_path = path_prefix + "/chinese_L-12_H-768_A-12/bert_config.json"
checkpoint_path = path_prefix + "/chinese_L-12_H-768_A-12/bert_model.ckpt"
dict_path = path_prefix + "/chinese_L-12_H-768_A-12/vocab.txt"
bert_model = load_trained_model_from_checkpoint(config_path, checkpoint_path, trainable=False)
x1_in = Input(shape=(None,))
x2_in = Input(shape=(None,))
x = bert_model([x1_in, x2_in])
x = Lambda(lambda x: x[:, 0])(x) # 取出[CLS]对应的向量用来做分类，LCS在第一位，然后才依次是其它序列的向量输出
x = Dropout(0.2)(x)
p = Dense(1, activation='sigmoid')(x)
model = Model(inputs=[x1_in, x2_in], outputs=p)
model.compile(
loss='binary_crossentropy',
optimizer=Adam(0.05),
metrics=['accuracy']
)
model.summary()
data = []
# 读取数据
read_data = open("./data/sentiment.test.data", encoding="utf-8").readlines()
for i in read_data:
a = i.replace("\n", "").split("\t")
if a[1] == "1":
data.append((a[0], 1))
else:
data.append((a[0], 0))
# 读取字典
token_dict = load_vocabulary(dict_path) # {word:id,word1:id1}
# 建立分词器
tokenizer = Tokenizer(token_dict)
#
# # 拆分单词实例
# # ids = tokenizer.tokenize("unaffable") # ['[CLS]', 'u', '##na', '##ff', '##able', '[SEP]']
# # a,b = tokenizer.encode(first="我不喜欢你",second="你也不喜欢我")
# # a:word对应的id.
# # b:segment对应的id
#
# # 按照9:1的比例划分训练集和验证集
random_order = list(range(len(data)))
np.random.shuffle(random_order)
train_data = [data[j] for i, j in enumerate(random_order) if i % 10 != 0]
valid_data = [data[j] for i, j in enumerate(random_order) if i % 10 == 0]
def seq_padding(X, padding=0):
L = [len(x) for x in X]
ML = max(L)
return np.array([
np.concatenate([x, [padding] * (ML - len(x))]) if len(x) < ML else x for x in X
])
class data_generator:
def __init__(self, data, batch_size=batch_size):
self.data = data
self.batch_size = batch_size
self.steps = len(self.data) // self.batch_size
if len(self.data) % self.batch_size != 0:
self.steps += 1
def __len__(self):
return self.steps
def __iter__(self):
while True:
idxs = list(range(len(self.data)))
np.random.shuffle(idxs)
X1, X2, Y = [], [], []
for i in idxs:
d = self.data[i]
text = d[0][:maxlen]
x1, x2 = tokenizer.encode(first=text)
y = d[1]
X1.append(x1)
X2.append(x2)
Y.append([y])
if len(X1) == self.batch_size or i == idxs[-1]:
X1 = seq_padding(X1)
X2 = seq_padding(X2)
Y = seq_padding(Y)
yield [X1, X2], Y
[X1, X2, Y] = [], [], []
# 相当于是batch_token_id, batch_segment_id, batch_labels
# train_D = data_generator(train_data)
# valid_D = data_generator(valid_data)
def get_data(data):
idxs = list(range(len(data)))
np.random.shuffle(idxs)
X1, X2, Y = [], [], []
for i in idxs:
d = data[i]
text = d[0]
x1, x2 = tokenizer.encode(first=text)
y = d[1]
x1 = np.concatenate([x1, [0] * (100 - len(x1))]) if len(x1) < 100 else x1[:100]
x2 = np.concatenate([x2, [0] * (100 - len(x2))]) if len(x2) < 100 else x2[:100]
# y = np.concatenate([y, [0] * (100 - len(y))]) if len(y) < 100 else y[:100]
X1.append(x1)
X2.append(x2)
Y.append(y)
return X1, X2, Y
X1, X2, Y = get_data(train_data)
print(Y)
model.fit([X1, X2], Y, batch_size=8,epochs=5)
# model.fit_generator(
# train_D.__iter__(),
# steps_per_epoch=len(train_D),
# epochs=epochs,
# validation_data=valid_D.__iter__(),
# validation_steps=len(valid_D)
# )
# model.save("new.model")
#
# for i in range(5):
# A, B = [], []
# a, b = tokenizer.encode(first=data[i][0])
# A.append(a)
# B.append(b)
# A = seq_padding(A)
# B = seq_padding(B)
# print(A)
# print(B)
# a = model.predict(x=[A, B])
# print(a, data[i][1])
# print("------------------------------")
#
# # pred = model.predict()
# # print(pred)
标签：bert,调用,keras,self,len,x2,model,X1,data 来源： https://blog.csdn.net/zhangjing_angry/article/details/114262565
本站声明： 1. iCode9 技术分享网（下文简称本站）提供的所有内容，仅供技术学习、探讨和分享； 2. 关于本站的所有留言、评论、转载及引用，纯属内容发起人的个人观点，与本站观点和立场无关； 3. 关于本站的所有言论和文字，纯属内容发起人的个人观点，与本站观点和立场无关； 4. 本站文章均是网友提供，不完全保证技术分享内容的完整性、准确性、时效性、风险性和版权归属；如您发现该文章侵犯了您的权益，可联系我们第一时间进行删除； 5. 本站为非盈利性的个人网站，所有内容不会用来进行牟利，也不会利用任何形式的广告来间接获益，纯粹是为了广大技术爱好者提供技术内容和技术思想的分享性交流网站。