从csv文件构建Tensorflow的数据集

从csv文件构建Tensorflow的数据集

当我们有一系列CSV文件，如何构建Tensorflow的数据集呢?

基本步骤

1. 获得一组CSV文件的路径
2. 将这组文件名，转成文件名对应的dataset => file_dataset
3. 根据file_dataset中的每个文件名，读取文件内容 生成一个内容的dataset => content_dataset
4. 这样的多个content_dataset, 拼接起来，形成一整个dataset
5. 因为读出来的每条记录都是string类型， 所以还需要对每条记录做decode

存在一个这样的变量train_filenames

pprint.pprint(train_filenames)
#	['generate_csv\\train_00.csv',
#	 'generate_csv\\train_01.csv',
#	 'generate_csv\\train_02.csv',
#	 'generate_csv\\train_03.csv',
#	 'generate_csv\\train_04.csv',
#	 'generate_csv\\train_05.csv',
#	 'generate_csv\\train_06.csv',
#	 'generate_csv\\train_07.csv',
#	 'generate_csv\\train_08.csv',
#	 'generate_csv\\train_09.csv',
#	 'generate_csv\\train_10.csv',
#	 'generate_csv\\train_11.csv',
#	 'generate_csv\\train_12.csv',
#	 'generate_csv\\train_13.csv',
#	 'generate_csv\\train_14.csv',
#	 'generate_csv\\train_15.csv',
#	 'generate_csv\\train_16.csv',
#	 'generate_csv\\train_17.csv',
#	 'generate_csv\\train_18.csv',
#	 'generate_csv\\train_19.csv']

接着，我们用提前定义好的API构建文件名数据集file_dataset

filename_dataset = tf.data.Dataset.list_files(train_filenames)
for filename in filename_dataset:
    print(filename)
#tf.Tensor(b'generate_csv\\train_09.csv', shape=(), dtype=string)
#tf.Tensor(b'generate_csv\\train_19.csv', shape=(), dtype=string)
#tf.Tensor(b'generate_csv\\train_03.csv', shape=(), dtype=string)
#tf.Tensor(b'generate_csv\\train_01.csv', shape=(), dtype=string)
#tf.Tensor(b'generate_csv\\train_14.csv', shape=(), dtype=string)
#tf.Tensor(b'generate_csv\\train_17.csv', shape=(), dtype=string)
#tf.Tensor(b'generate_csv\\train_15.csv', shape=(), dtype=string)
#tf.Tensor(b'generate_csv\\train_06.csv', shape=(), dtype=string)
#tf.Tensor(b'generate_csv\\train_05.csv', shape=(), dtype=string)
#tf.Tensor(b'generate_csv\\train_07.csv', shape=(), dtype=string)
#tf.Tensor(b'generate_csv\\train_11.csv', shape=(), dtype=string)
#tf.Tensor(b'generate_csv\\train_02.csv', shape=(), dtype=string)
#tf.Tensor(b'generate_csv\\train_12.csv', shape=(), dtype=string)
#tf.Tensor(b'generate_csv\\train_13.csv', shape=(), dtype=string)
#tf.Tensor(b'generate_csv\\train_10.csv', shape=(), dtype=string)
#tf.Tensor(b'generate_csv\\train_16.csv', shape=(), dtype=string)
#tf.Tensor(b'generate_csv\\train_18.csv', shape=(), dtype=string)
#tf.Tensor(b'generate_csv\\train_00.csv', shape=(), dtype=string)
#tf.Tensor(b'generate_csv\\train_04.csv', shape=(), dtype=string)
#tf.Tensor(b'generate_csv\\train_08.csv', shape=(), dtype=string)

第三步， 根据每个文件名，去读取文件里面的内容

dataset = filename_dataset.interleave(
    lambda filename: tf.data.TextLineDataset(filename).skip(1),
    cycle_length=5
)

for line in dataset.take(3):
    print(line)

#tf.Tensor(b'0.46908349737250216,1.8718193706428006,0.13936365871212536,-0.011055733363841472,-0.6349261778219746,-0.036732316700563934,1.0259470089944995,-1.319095600336748,2.171', shape=(), dtype=string)
#tf.Tensor(b'-1.102093775650278,1.313248890578542,-0.7212003024178728,-0.14707856286537277,0.34720121604358517,0.0965085401826684,-0.74698820254838,0.6810563907247876,1.428', shape=(), dtype=string)
#tf.Tensor(b'-0.8901003715328659,0.9142699762469286,-0.1851678950250224,-0.12947457252940406,0.5958187430364827,-0.021255215877779534,0.7914317693724252,-0.45618713536506217,0.75', shape=(), dtype=string)

interleave的作用可以类比map, 对每个元素应用操作，然后还能把结果合起来。
因此，有了interleave, 我们就把第三四步，一起完成了
之所以skip(1),是因为这个csv第一行是header.
cycle_length是并行化构建数据集的线程数

好，第五步，解析每条记录

def parse_csv_line(line, n_fields=9):
    defaults = [tf.constant(np.nan)] * n_fields
    parsed_fields = tf.io.decode_csv(line, record_defaults=defaults)
    x = tf.stack(parsed_fields[:-1])
    y = tf.stack(parsed_fields[-1:])
    return x, y

parse_csv_line('1.2286258796252256,-1.0806245954111382,0.4444161407754224,-0.0352172575329119,0.9740347681426992,-0.003516079473801425,-0.8126524696425611,0.865609068204283,2.803', 9)

#(<tf.Tensor: shape=(8,), dtype=float32, numpy= array([ 1.2286259 , -1.0806246 ,  0.44441614, -0.03521726,  0.9740348 ,-0.00351608, -0.81265247,  0.86560905], dtype=float32)>,<tf.Tensor: shape=(1,), dtype=float32, numpy=array([2.803], dtype=float32)>)

最后，将每条记录都应用这个方法，就完成了构建。

dataset = dataset.map(parse_csv_line)

完整代码

def csv_2_dataset(filenames, n_readers_thread = 5, batch_size = 32, n_parse_thread = 5, shuffle_buffer_size = 10000):
    
    dataset = tf.data.Dataset.list_files(filenames)
    dataset = dataset.repeat()
    dataset = dataset.interleave(
        lambda filename: tf.data.TextLineDataset(filename).skip(1),
        cycle_length=n_readers_thread
    )
    dataset.shuffle(shuffle_buffer_size)
    dataset = dataset.map(parse_csv_line, num_parallel_calls = n_parse_thread)
    dataset = dataset.batch(batch_size)
    return dataset

如何使用

train_dataset = csv_2_dataset(train_filenames, batch_size=32)
valid_dataset = csv_2_dataset(valid_filenames, batch_size=32)

model = ...

model.fit(train_set, validation_data=valid_set, 
                   steps_per_epoch = 11610 // 32,
                   validation_steps = 3870 // 32,
                   epochs=100, callbacks=callbacks)

这里的11610 和 3870是什么？

这是train_dataset 和 valid_dataset中数据的数量，需要在训练中手动指定每个batch中参与训练的数据的多少。

model.evaluate(test_set, steps=5160//32)

同理，测试的时候，使用这样的数据集，也需要手动指定。
5160是测试数据集的总量。