[深度学习-实战]LSTM Attention训练IMDB文本的情感分析_lstm-attention情感分析

Train.npz， test.npz 以及embedding_matrix.npy的产生请看这本博客

完整代码如下

import os
import numpy as np
import tensorflow.keras as keras
import tensorflow.keras.layers as layers
import tensorflow as tf
import time

root_folder = './lstm_attention'
def get_dataset():
    train_set = np.load('./train_data_new1/train.npz')
    X_train = train_set['x']
    y_train = train_set['y']
    test_set = np.load('./train_data_new1/test.npz')
    X_test = test_set['x']
    y_test = test_set['y']

    print("X_train:", X_train.shape)
    print("y_train:", y_train.shape)
    print("X_test:", X_test.shape)
    print("y_test:", y_test.shape)
    return X_train, y_train, X_test, y_test

def lstm_model():
    embedding_matrix = np.load('./train_data_new1/embedding_matrix.npy')

    input = tf.keras.Input(batch_shape=(None, 200))
    embed = tf.keras.layers.Embedding(input_dim=30001, output_dim=50, weights=[embedding_matrix],trainable=True)(input)
    bilstm = tf.keras.layers.Bidirectional(tf.keras.layers.LSTM(128, return_sequences=True, dropout=0.5))(embed)
    attention_layer = tf.keras.layers.Attention()([bilstm, bilstm])
    pooling_out1 = tf.keras.layers.GlobalMaxPooling1D()(bilstm)
    pooling_out2 = tf.keras.layers.GlobalMaxPooling1D()(attention_layer)
    merge = tf.keras.layers.Concatenate()([pooling_out1, pooling_out2])
    mlp_hidden_out = tf.keras.layers.Dense(64, activation=tf.nn.relu)(merge)
    out = tf.keras.layers.Dense(2, activation=tf.nn.softmax)(mlp_hidden_out)

    model = tf.keras.Model(inputs=[input], outputs=[out])
    model.compile(optimizer='adam',
                  loss='sparse_categorical_crossentropy',
                  metrics=['accuracy'])
    model.summary()
    return model

current_max_loss =9999
def train_my_model(model, X_train, y_train):
    weight_dir = root_folder+'\weight.h5'

    if os.path.isfile(weight_dir):
        print('load weight')
        model.load_weights(weight_dir)

    def save_weight(epoch, logs):
        global current_max_loss
        if(logs['val_loss'] is not None and  logs['val_loss']< current_max_loss):
            current_max_loss = logs['val_loss']
            print('save_weight', epoch, current_max_loss)
            model.save_weights(weight_dir)
            model.save(root_folder+'\model.h5', include_optimizer=False)

    batch_print_callback = keras.callbacks.LambdaCallback(
        on_epoch_end=save_weight
    )
    callbacks = [
        tf.keras.callbacks.EarlyStopping(patience=4, monitor='loss'),
        batch_print_callback,
        tf.keras.callbacks.TensorBoard(log_dir=root_folder+'\logs')
    ]
    begin = time.time()
    history = model.fit(X_train, y_train, batch_size=128, epochs=25,validation_split=0.1, callbacks= callbacks)
    finish = time.time()
    print("train time: ", (finish - begin), 's')
    import matplotlib.pyplot as plt
    plt.plot(history.history['accuracy'])
    plt.plot(history.history['val_accuracy'])
    plt.legend(['accuracy', 'val_accuracy'], loc='upper left')
    plt.show()

def test_my_module(model, X_test, y_test):
    if os.path.isfile(root_folder+'/model.h5'):
        print('load weight')
        model.load_weights(root_folder+'/model.h5')
    test_result = model.evaluate(X_test, y_test)
    print('test Result', test_result)
    print('Test ',test_result)


def predict_my_module(model):
    small_word_index = np.load('./train_data_new1/small_word_index.npy', allow_pickle=True)

    review_index = np.zeros((1, 200), dtype=int)
    #review = "I don't like it"
    #review = "this is bad movie "
    #review = "This is good movie"
    #review = "This isn‘t great movie"
    #review = "i think this is bad movie"
    #review = "I'm not very disappoint for this movie"
    #review = "I'm not very disappoint for this movie"
    review = "I'm very happy for this movie"
    #neg:0 postive:1
    counter = 0
    for word in review.split():
        try:
            print(word, small_word_index.item()[word])
            review_index[0][counter] = small_word_index.item()[word]
            counter = counter + 1
        except Exception:
            print('Word error', word)
    print(review_index.shape)
    s = model.predict(x=review_index)
    print(s)


if __name__ == '__main__':
    X_train, y_train, x_test, y_test = get_dataset()
    model = lstm_model()
    train_my_model(model, X_train, y_train)
    test_my_module(model,x_test, y_test)
    #predict_my_module(model)

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

测试集上的结果是 90.78%

15000/15000 [==============================] - 5s 340us/sample - loss: 0.2375 - accuracy: 0.9078
test Result [0.2374931192914645, 0.9078]
Test  [0.2374931192914645, 0.9078]
1
2
3

训练过程中损失函数以及准确率的变化曲线如下
在训练后，以通过下面命令看
tensorboard --logdir=logs