资源简介

tensorflow2.0的Lstm实现

资源截图

代码片段和文件信息

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


tf.random.set_seed(22)
np.random.seed(22)
os.environ[‘TF_CPP_MIN_LOG_LEVEL‘] = ‘2‘
assert tf.__version__.startswith(‘2.‘)

batchsz = 128

# the most frequest words
total_words = 10000
max_review_len = 80
embedding_len = 100
(x_train y_train) (x_test y_test) = keras.datasets.imdb.load_data(num_words=total_words)
# x_train:[b 80]
# x_test: [b 80]
x_train = keras.preprocessing.sequence.pad_sequences(x_train maxlen=max_review_len)
x_test = keras.preprocessing.sequence.pad_sequences(x_test maxlen=max_review_len)

db_train = tf.data.Dataset.from_tensor_slices((x_train y_train))
db_train = db_train.shuffle(1000).batch(batchsz drop_remainder=True)
db_test = tf.data.Dataset.from_tensor_slices((x_test y_test))
db_test = db_test.batch(batchsz drop_remainder=True)
print(‘x_train shape:‘ x_train.shape tf.reduce_max(y_train) tf.reduce_min(y_train))
print(‘x_test shape:‘ x_test.shape)



class MyRNN(keras.Model):

    def __init__(self units):
        super(MyRNN self).__init__()

        # [b 64]
        self.state0 = [tf.zeros([batchsz units])tf.zeros([batchsz units])]
        self.state1 = [tf.zeros([batchsz units])tf.zeros([batchsz units])]

        # transform text to embedding representation
        # [b 80] => [b 80 100]
        self.embedding = layers.embedding(total_words embedding_len
                                          input_length=max_review_len)

        # [b 80 100]  h_dim: 64
        # RNN: cell1 cell2 cell3
        # SimpleRNN
        # self.rnn_cell0 = layers.SimpleRNNCell(units dropout=0.5)
        # self.rnn_cell1 = layers.SimpleRNNCell(units dropout=0.5)
        self.rnn_cell0 = layers.LSTMCell(units dropout=0.5)
        self.rnn_cell1 = layers.LSTMCell(units dropout=0.5)


        # fc [b 80 100] => [b 64] => [b 1]
        self.outlayer = layers.Dense(1)

    def call(self inputs training=None):
        “““
        net(x) net(x training=True) :train mode
        net(x training=False): test
        :param inputs: [b 80]
        :param training:
        :return:
        “““
        # [b 80]
        x = inputs
        # embedding: [b 80] => [b 80 100]
        x = self.embedding(x)
        # rnn cell compute
        # [b 80 100] => [b 64]
        state0 = self.state0
        state1 = self.state1
        for word in tf.unstack(x axis=1): # word: [b 100]
            # h1 = x*wxh+h0*whh
            # out0: [b 64]
            out0 state0 = self.rnn_cell0(word state0 training)
            # out1: [b 64]
            out1 state1 = self.rnn_cell1(out0 state1 training)

        # out: [b 64] => [b 1]
        x = self.outlayer(out1)
        # p(y is pos|x)
        prob = tf.sigmoid(x)

        return prob

def main():
    units = 64
    epochs = 4

    import time

    t0 = time.time()

    model = MyRNN(units)
    model.compile(optimizer = keras.optimizer

 属性            大小     日期    时间   名称
----------- ---------  ---------- -----  ----

     文件       2868  2020-02-16 15:57  lstm_sentiment_analysis_layer.py

     文件       3330  2020-02-16 15:57  lstm_sentiment_analysis_cell.py

----------- ---------  ---------- -----  ----

                 6198                    2


评论

共有 条评论