TF之RNN:实现利用scope.reuse_variables()告诉TF想重复利用RNN的参数的案例—Jason niu
import tensorflow as tf
# 22 scope (name_scope/variable_scope)
from __future__ import print_function class TrainConfig:
batch_size = 20
time_steps = 20
input_size = 10
output_size = 2
cell_size = 11
learning_rate = 0.01 class TestConfig(TrainConfig):
time_steps = 1 class RNN(object): def __init__(self, config):
self._batch_size = config.batch_size
self._time_steps = config.time_steps
self._input_size = config.input_size
self._output_size = config.output_size
self._cell_size = config.cell_size
self._lr = config.learning_rate
self._built_RNN() def _built_RNN(self):
with tf.variable_scope('inputs'):
self._xs = tf.placeholder(tf.float32, [self._batch_size, self._time_steps, self._input_size], name='xs')
self._ys = tf.placeholder(tf.float32, [self._batch_size, self._time_steps, self._output_size], name='ys')
with tf.name_scope('RNN'):
with tf.variable_scope('input_layer'):
l_in_x = tf.reshape(self._xs, [-1, self._input_size], name='2_2D') # (batch*n_step, in_size)
# Ws (in_size, cell_size)
Wi = self._weight_variable([self._input_size, self._cell_size])
print(Wi.name)
# bs (cell_size, )
bi = self._bias_variable([self._cell_size, ])
# l_in_y = (batch * n_steps, cell_size)
with tf.name_scope('Wx_plus_b'):
l_in_y = tf.matmul(l_in_x, Wi) + bi
l_in_y = tf.reshape(l_in_y, [-1, self._time_steps, self._cell_size], name='2_3D') with tf.variable_scope('cell'):
cell = tf.contrib.rnn.BasicLSTMCell(self._cell_size)
with tf.name_scope('initial_state'):
self._cell_initial_state = cell.zero_state(self._batch_size, dtype=tf.float32) self.cell_outputs = []
cell_state = self._cell_initial_state
for t in range(self._time_steps):
if t > 0: tf.get_variable_scope().reuse_variables()
cell_output, cell_state = cell(l_in_y[:, t, :], cell_state)
self.cell_outputs.append(cell_output)
self._cell_final_state = cell_state with tf.variable_scope('output_layer'):
# cell_outputs_reshaped (BATCH*TIME_STEP, CELL_SIZE)
cell_outputs_reshaped = tf.reshape(tf.concat(self.cell_outputs, 1), [-1, self._cell_size])
Wo = self._weight_variable((self._cell_size, self._output_size))
bo = self._bias_variable((self._output_size,))
product = tf.matmul(cell_outputs_reshaped, Wo) + bo
# _pred shape (batch*time_step, output_size)
self._pred = tf.nn.relu(product) # for displacement with tf.name_scope('cost'):
_pred = tf.reshape(self._pred, [self._batch_size, self._time_steps, self._output_size])
mse = self.ms_error(_pred, self._ys)
mse_ave_across_batch = tf.reduce_mean(mse, 0)
mse_sum_across_time = tf.reduce_sum(mse_ave_across_batch, 0)
self._cost = mse_sum_across_time
self._cost_ave_time = self._cost / self._time_steps with tf.variable_scope('trian'):
self._lr = tf.convert_to_tensor(self._lr)
self.train_op = tf.train.AdamOptimizer(self._lr).minimize(self._cost) @staticmethod
def ms_error(y_target, y_pre):
return tf.square(tf.subtract(y_target, y_pre)) @staticmethod
def _weight_variable(shape, name='weights'):
initializer = tf.random_normal_initializer(mean=0., stddev=0.5, )
return tf.get_variable(shape=shape, initializer=initializer, name=name) @staticmethod
def _bias_variable(shape, name='biases'):
initializer = tf.constant_initializer(0.1)
return tf.get_variable(name=name, shape=shape, initializer=initializer) if __name__ == '__main__':
train_config = TrainConfig() #定义train_config
test_config = TestConfig() # # the wrong method to reuse parameters in train rnn
# with tf.variable_scope('train_rnn'):
# train_rnn1 = RNN(train_config)
# with tf.variable_scope('test_rnn'):
# test_rnn1 = RNN(test_config) # the right method to reuse parameters in train rnn
#目的使train的RNN调用参数,然后利用variable_scope方法共享RNN,让test的RNN再次调用一样的参数,
with tf.variable_scope('rnn') as scope:
sess = tf.Session()
train_rnn2 = RNN(train_config)
scope.reuse_variables() #告诉TF想重复利用RNN的参数
test_rnn2 = RNN(test_config)
# tf.initialize_all_variables() no long valid from
# 2017-03-02 if using tensorflow >= 0.12
if int((tf.__version__).split('.')[1]) < 12 and int((tf.__version__).split('.')[0]) < 1:
init = tf.initialize_all_variables()
else:
init = tf.global_variables_initializer()
sess.run(init)
TF之RNN:实现利用scope.reuse_variables()告诉TF想重复利用RNN的参数的案例—Jason niu的更多相关文章
- TF之RNN:TF的RNN中的常用的两种定义scope的方式get_variable和Variable—Jason niu
# tensorflow中的两种定义scope(命名变量)的方式tf.get_variable和tf.Variable.Tensorflow当中有两种途径生成变量 variable import te ...
- 深度学习原理与框架-递归神经网络-RNN_exmaple(代码) 1.rnn.BasicLSTMCell(构造基本网络) 2.tf.nn.dynamic_rnn(执行rnn网络) 3.tf.expand_dim(增加输入数据的维度) 4.tf.tile(在某个维度上按照倍数进行平铺迭代) 5.tf.squeeze(去除维度上为1的维度)
1. rnn.BasicLSTMCell(num_hidden) # 构造单层的lstm网络结构 参数说明:num_hidden表示隐藏层的个数 2.tf.nn.dynamic_rnn(cell, ...
- TF之RNN:matplotlib动态演示之基于顺序的RNN回归案例实现高效学习逐步逼近余弦曲线—Jason niu
import tensorflow as tf import numpy as np import matplotlib.pyplot as plt BATCH_START = 0 TIME_STEP ...
- TF之RNN:基于顺序的RNN分类案例对手写数字图片mnist数据集实现高精度预测—Jason niu
import tensorflow as tf from tensorflow.examples.tutorials.mnist import input_data mnist = input_dat ...
- TensorFlow RNN MNIST字符识别演示快速了解TF RNN核心框架
TensorFlow RNN MNIST字符识别演示快速了解TF RNN核心框架 http://blog.sina.com.cn/s/blog_4b0020f30102wv4l.html
- TF之RNN:TensorBoard可视化之基于顺序的RNN回归案例实现蓝色正弦虚线预测红色余弦实线—Jason niu
import tensorflow as tf import numpy as np import matplotlib.pyplot as plt BATCH_START = 0 TIME_STEP ...
- TF:利用sklearn自带数据集使用dropout解决学习中overfitting的问题+Tensorboard显示变化曲线—Jason niu
import tensorflow as tf from sklearn.datasets import load_digits #from sklearn.cross_validation impo ...
- TF:Tensorflow结构简单应用,随机生成100个数,利用Tensorflow训练使其逼近已知线性直线的效率和截距—Jason niu
import os os.environ[' import tensorflow as tf import numpy as np x_data = np.random.rand(100).astyp ...
- 深度学习原理与框架-图像补全(原理与代码) 1.tf.nn.moments(求平均值和标准差) 2.tf.control_dependencies(先执行内部操作) 3.tf.cond(判别执行前或后函数) 4.tf.nn.atrous_conv2d 5.tf.nn.conv2d_transpose(反卷积) 7.tf.train.get_checkpoint_state(判断sess是否存在
1. tf.nn.moments(x, axes=[0, 1, 2]) # 对前三个维度求平均值和标准差,结果为最后一个维度,即对每个feature_map求平均值和标准差 参数说明:x为输入的fe ...
随机推荐
- Ajax中onreadystatechange函数不执行,是因为放在open()后
今天动手写Ajax时遇到的问题:按照下面的顺序来写Ajax,功能是alert出txt文档的内容,在Chrome下可以执行onreadystatechange中的事件,在IE11及以下.FF就不能执行o ...
- SecureCRT中sqlplus,使用Backspace删除时 ^H^H
平时习惯用Backspace删除输入错误,但是在SecureCRT中使用是,却是: SQL> sele^H^H 网上有几个方法,觉得改SecureCRT的配置最方便.
- Confluence 6 发送 Confluence 通知到其他 Confluence 服务器
你可以配置 Confluence 服务器向其他的 Confluence 服务器发送消息.在这种情况下,Confluence 服务器将不会显示 workbox. 希望发送消息到其他 Confluence ...
- Burp Scanner Report
1.使用application web 漏洞平台,除此之外还有一款类似的工具 叫做mulidata,其实mulidata功能更好一点. 2.配置之前的问题处理 安装之前要确认 自己之前是否安装过 Ap ...
- 2018.8.1 状压 CF482C 题解
noip2016考了一道状压dp,一道期望dp 然而这题是状压期望dp... 所以难度是什么,省选noi吗... 怎么办... 题目大意: 给定n个字符串,甲从中任选出一个串(即选出每个串的概率相同为 ...
- spring cloud 使用ribbon简单处理客户端负载均衡
假如我们的multiple服务的访问量剧增,用一个服务已经无法承载, 我们可以把Hello World服务做成一个集群. 很简单,我们只需要复制Hello world服务,同时将原来的端口8762修改 ...
- Python零基础入门之Tkinter的对话框
这篇博客主要是总结一下Tkinter中的对话框的使用,值得一提的是自从python3.0之后关于关于对话框的模块(messagebox.filedialog.colorchooser)都被收归到了tk ...
- IDM的Google商店插件
官方扩展链接:https://chrome.google.com/webstore/detail/idm-integration-module/ngpampappnmepgilojfohadhhmbh ...
- WCF三种通信方式
一.概述 WCF在通信过程中有三种模式:请求与答复.单向.双工通信.以下我们一一介绍. 二.请求与答复模式 描述: 客户端发送请求,然后一直等待服务端的响应(异步调用除外),期间处于假死状态,直到服务 ...
- C# 属性(Property)和字段(Field)的区别
导读: 近期学习过程中发现了一些问题,我的学习只是学习,敲代码就是敲代码,没有加入思考,也不问为什么就直接去敲人家写好的例子去敲,把知识都学死了,逐渐散失了思考能力,所以学习的兴趣大打折扣,正如那句话 ...