流程

先介紹下數(shù)據(jù)和整體流程

生成數(shù)據(jù)

import numpy as nprows = 10000 columns = 100 emb_size = 5train_x = np.random.random(size=(rows, columns, emb_size)) train_y = np.random.randint(low=0, high=2, size=(rows, 1))

tf 計(jì)算流程

import tensorflow as tf from tensorflow import kerasmodel = keras.Sequential() model.add(keras.layers.Input(shape=[columns, emb_size], name='my_input')) # batch_size為None model.add(keras.layers.SimpleRNN(units=10, use_bias=False, kernel_initializer=keras.initializers.Zeros())) model.add(keras.layers.Dense(1, kernel_initializer=keras.initializers.Orthogonal(gain=1.0, seed=None)))model.compile(loss=tf.keras.losses.BinaryCrossentropy(from_logits=True),optimizer=tf.keras.optimizers.Adam(1e-4),metrics=['accuracy'])model.fit(train_x, train_y, epochs=10, batch_size=100)

tf 實(shí)現(xiàn)SimpleRNN

class MySimpleRNN(keras.layers.B):def __init__(self):super(MySimpleRNN, self).__init__()passdef build(self):passdef call(self):pass

numpy 計(jì)算流程

import numpy as np

總結(jié)

以上是生活随笔為你收集整理的tf rnn layer的全部內(nèi)容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網(wǎng)站內(nèi)容還不錯(cuò)，歡迎將生活随笔推薦給好友。