tensorflow 批量读取csv文件用于做深度学习算法相关

目前用了tensorflow、deeplearning4j两个深度学习框架，dl相关算法对数据格式要求都是批量的喂进去，deepl4j在前面已经有几个例子说明，tensorflow也可以批量读取数据，不断给dl算法喂数据进去，在网上刚刚看到一个例子，http://www.cnblogs.com/hunttown/p/6844477.html ，首先数据格式如下，鸾尾花数据
做机器学习的人应该都知道：

1. Id,SepalLengthCm,SepalWidthCm,PetalLengthCm,PetalWidthCm,Species
2. 21,5.4,3.4,1.7,0.2,Iris-setosa
3. 22,5.1,3.7,1.5,0.4,Iris-setosa
4. 23,4.6,3.6,1.0,0.2,Iris-setosa
5. 24,5.1,3.3,1.7,0.5,Iris-setosa
6. 25,4.8,3.4,1.9,0.2,Iris-setosa
7. 26,5.0,3.0,1.6,0.2,Iris-setosa
8. 27,5.0,3.4,1.6,0.4,Iris-setosa
9. 28,5.2,3.5,1.5,0.2,Iris-setosa
10. 29,5.2,3.4,1.4,0.2,Iris-setosa
11. 30,4.7,3.2,1.6,0.2,Iris-setosa
12. 31,4.8,3.1,1.6,0.2,Iris-setosa
13. 32,5.4,3.4,1.5,0.4,Iris-setosa
14. 33,5.2,4.1,1.5,0.1,Iris-setosa
15. 34,5.5,4.2,1.4,0.2,Iris-setosa
16. 35,4.9,3.1,1.5,0.1,Iris-setosa
17. 36,5.0,3.2,1.2,0.2,Iris-setosa
18. 37,5.5,3.5,1.3,0.2,Iris-setosa
19. 39,5.5,4.2,1.4,0.2,Iris-virginica
20. 40,4.9,3.1,1.5,0.1,Iris-versicolor
21. 38,5.0,3.2,1.2,0.2,Iris-versicolor
22. 51,5.5,3.5,1.3,0.2,Iris-versicolor

下面是程序实现：

1. import tensorflow as tf
2. path="/Users/shuubiasahi/Desktop/业务相关文档/iris.csv"
4. def read_data(file_queue):
5. reader=tf.TextLineReader(skip_header_lines=1)
6. key,value=reader.read(file_queue)
7. defaults=[[0], [0.], [0.], [0.], [0.], ['']]
8. Id,SepalLengthCm,SepalWidthCm,PetalLengthCm,PetalWidthCm,Species = tf.decode_csv(value, defaults)
11. preprocess_op=tf.case({
12. tf.equal(Species,tf.constant('Iris-setosa')):lambda :tf.constant(0),
13. tf.equal(Species, tf.constant('Iris-versicolor')): lambda: tf.constant(1),
14. tf.equal(Species, tf.constant('Iris-virginica')): lambda: tf.constant(2),
16. },lambda :tf.constant(-1),exclusive=True)
18. return tf.stack([SepalLengthCm, SepalWidthCm, PetalLengthCm, PetalWidthCm]), preprocess_op
20. def create_pipeline(filename,batch_size,num_epochs=None):
21. file_queue = tf.train.string_input_producer([filename], num_epochs=num_epochs)
22. example, label = read_data(file_queue)
23. min_after_dequeue = 1000
24. capacity = min_after_dequeue + batch_size
25. example_batch, label_batch = tf.train.shuffle_batch(
26. [example, label], batch_size=batch_size, capacity=capacity,
27. min_after_dequeue=min_after_dequeue
28. )
30. return example_batch, label_batch
32. x_train_batch, y_train_batch = create_pipeline(path, 5, num_epochs=1000)
33. x_test, y_test = create_pipeline(path, 60)
34. init_op = tf.global_variables_initializer()
35. local_init_op = tf.local_variables_initializer() # local variables like epoch_num, batch_size
36. with tf.Session() as sess:
37. sess.run(init_op)
38. sess.run(local_init_op)
39. # Start populating the filename queue.
40. coord = tf.train.Coordinator()
41. threads = tf.train.start_queue_runners(coord=coord)
43. # Retrieve a single instance:
44. try:
45. #while not coord.should_stop():
46. for _ in range(6):
47. example, label = sess.run([x_test, y_test])
48. print (example)
49. print (label)
50. except tf.errors.OutOfRangeError:
51. print ('Done reading')
52. finally:
53. coord.request_stop()
54. coord.join(threads)
55. sess.close()