[Jupyter] Learning TensorFlow

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

本帖隐藏的内容

https://github.com/Hezi-Resheff/Oreilly-Learning-TensorFlow

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏1 回帖

关键词：Learning earning Tensor Learn flow

1. from tensorflow.examples.tutorials.mnist import input_data
   
2. import tensorflow as tf
   
3. import numpy as np
   
4. 
   
5. DATA_DIR = '/tmp/data'
   
6. MINIBATCH_SIZE = 50
   
7. STEPS = 5000
   
8. 
   
9. from layers import *
   
10. 
    
11. mnist = input_data.read_data_sets(DATA_DIR, one_hot=True)
    
12. 
    
13. x = tf.placeholder(tf.float32, shape=[None, 784])
    
14. y_ = tf.placeholder(tf.float32, shape=[None, 10])
    
15. 
    
16. x_image = tf.reshape(x, [-1, 28, 28, 1])
    
17. conv1 = conv_layer(x_image, shape=[5, 5, 1, 32])
    
18. conv1_pool = max_pool_2x2(conv1)
    
19. 
    
20. conv2 = conv_layer(conv1_pool, shape=[5, 5, 32, 64])
    
21. conv2_pool = max_pool_2x2(conv2)
    
22. 
    
23. conv2_flat = tf.reshape(conv2_pool, [-1, 7*7*64])
    
24. full_1 = tf.nn.relu(full_layer(conv2_flat, 1024))
    
25. 
    
26. keep_prob = tf.placeholder(tf.float32)
    
27. full1_drop = tf.nn.dropout(full_1, keep_prob=keep_prob)
    
28. 
    
29. y_conv = full_layer(full1_drop, 10)
    
30. 
    
31. cross_entropy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=y_conv, labels=y_))
    
32. train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
    
33. correct_prediction = tf.equal(tf.argmax(y_conv, 1), tf.argmax(y_, 1))
    
34. accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
    
35. 
    
36. with tf.Session() as sess:
    
37.     sess.run(tf.global_variables_initializer())
    
38. 
    
39.     for i in range(STEPS):
    
40.         batch = mnist.train.next_batch(MINIBATCH_SIZE)
    
41. 
    
42.         if i % 100 == 0:
    
43.             train_accuracy = sess.run(accuracy, feed_dict={x: batch[0], y_: batch[1],
    
44.                                                            keep_prob: 1.0})
    
45.             print("step {}, training accuracy {}".format(i, train_accuracy))
    
46. 
    
47.         sess.run(train_step, feed_dict={x: batch[0], y_: batch[1], keep_prob: 0.5})
    
48. 
    
49.     X = mnist.test.images.reshape(10, 1000, 784)
    
50.     Y = mnist.test.labels.reshape(10, 1000, 10)
    
51.     test_accuracy = np.mean([sess.run(accuracy, feed_dict={x:X[i], y_:Y[i], keep_prob:1.0}) for i in range(10)])
    
52. 
    
53. print("test accuracy: {}".format(test_accuracy))

复制代码

1. import tensorflow as tf
   
2. 
   
3. 
   
4. def weight_variable(shape):
   
5.     initial = tf.truncated_normal(shape, stddev=0.1)
   
6.     return tf.Variable(initial)
   
7. 
   
8. 
   
9. def bias_variable(shape):
   
10.     initial = tf.constant(0.1, shape=shape)
    
11.     return tf.Variable(initial)
    
12. 
    
13. 
    
14. def conv2d(x, W):
    
15.     return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding='SAME')
    
16. 
    
17. 
    
18. def max_pool_2x2(x):
    
19.     return tf.nn.max_pool(x, ksize=[1, 2, 2, 1],
    
20.                           strides=[1, 2, 2, 1], padding='SAME')
    
21. 
    
22. 
    
23. def conv_layer(input, shape):
    
24.     W = weight_variable(shape)
    
25.     b = bias_variable([shape[3]])
    
26.     return tf.nn.relu(conv2d(input, W) + b)
    
27. 
    
28. 
    
29. def full_layer(input, size):
    
30.     in_size = int(input.get_shape()[1])
    
31.     W = weight_variable([in_size, size])
    
32.     b = bias_variable([size])
    
33.     return tf.matmul(input, W) + b

复制代码

1. # -*- coding: utf-8 -*-
   
2. """
   
3. Created on Tue Dec 20 17:34:43 2016
   
4. 
   
5. @author: tomhope
   
6. """
   
7. from __future__ import print_function
   
8. import tensorflow as tf
   
9. from tensorflow.examples.tutorials.mnist import input_data
   
10. mnist = input_data.read_data_sets("/tmp/data/", one_hot=True)
    
11. 
    
12. element_size = 28;time_steps = 28;num_classes = 10
    
13. batch_size = 128;hidden_layer_size = 128
    
14. 
    
15. _inputs = tf.placeholder(tf.float32,shape=[None, time_steps,
    
16.                                            element_size],
    
17.                                            name='inputs')
    
18. y = tf.placeholder(tf.float32, shape=[None, num_classes],name='inputs')
    
19. 
    
20. #TensorFlow built-in functions
    
21. rnn_cell = tf.contrib.rnn.BasicRNNCell(hidden_layer_size)
    
22. outputs, _ = tf.nn.dynamic_rnn(rnn_cell, _inputs, dtype=tf.float32)
    
23. 
    
24. Wl = tf.Variable(tf.truncated_normal([hidden_layer_size, num_classes],
    
25.                                      mean=0,stddev=.01))
    
26. bl = tf.Variable(tf.truncated_normal([num_classes],mean=0,stddev=.01))
    
27. 
    
28. def get_linear_layer(vector):
    
29.     return tf.matmul(vector, Wl) + bl
    
30. 
    
31. last_rnn_output = outputs[:,-1,:]
    
32. final_output = get_linear_layer(last_rnn_output)
    
33. 
    
34. softmax = tf.nn.softmax_cross_entropy_with_logits(logits=final_output, labels=y)
    
35. cross_entropy = tf.reduce_mean(softmax)
    
36. train_step = tf.train.RMSPropOptimizer(0.001, 0.9).minimize(cross_entropy)
    
37. 
    
38. correct_prediction = tf.equal(tf.argmax(y,1), tf.argmax(final_output,1))
    
39. accuracy = (tf.reduce_mean(tf.cast(correct_prediction, tf.float32)))*100
    
40. 
    
41. sess=tf.InteractiveSession()
    
42. sess.run(tf.global_variables_initializer())
    
43. 
    
44. test_data = mnist.test.images[:batch_size].reshape((-1,
    
45.                                             time_steps, element_size))
    
46. test_label = mnist.test.labels[:batch_size]
    
47.   
    
48. for i in range(3001):
    
49. 
    
50.        batch_x, batch_y = mnist.train.next_batch(batch_size)
    
51.        batch_x = batch_x.reshape((batch_size, time_steps, element_size))
    
52.        sess.run(train_step,feed_dict={_inputs:batch_x,
    
53.                                       y:batch_y})
    
54.        if i % 1000 == 0:
    
55.             acc = sess.run(accuracy, feed_dict={_inputs: batch_x,
    
56.                                                 y: batch_y})
    
57.             loss = sess.run(cross_entropy,feed_dict={_inputs:batch_x,
    
58.                                                      y:batch_y})
    
59.             print ("Iter " + str(i) + ", Minibatch Loss= " + \
    
60.                   "{:.6f}".format(loss) + ", Training Accuracy= " + \
    
61.                   "{:.5f}".format(acc))   
    
62. 
    
63. print ("Testing Accuracy:", \
    
64.     sess.run(accuracy, feed_dict={_inputs: test_data, y: test_label}))
    
65.    

复制代码

ReneeBK 发表于 2017-8-13 06:48
**** 本内容被作者隐藏 ****

写字的分享

谢谢楼主分享！

谢谢楼主分享！

谢谢你的书！

谢谢分享