如何在训练和推理中使用tf.Dataset设计？ [英] How to use tf.Dataset design in both training and inferring?

问题描述

说，我们输入了 x 并标记了 y ：

Say, we have input x and label y:

iterator = tf.data.Iterator.from_structure((x_type, y_type), (x_shape, y_shape))
tf_x, tf_y = iterator.get_next()

现在我使用generate函数创建数据集：

Now I use generate function to create dataset:

def gen():
    for ....: yield (x, y)
ds = tf.data.Dataset.from_generator(gen, (x_type, y_type), (x_shape, y_shape))

在我的图形中，我使用 tf_x 和 tf_y 进行培训，很好。但是，现在我想引用，在这里我没有标签 y 。我采取的一种解决方法是伪造ay（例如tf.zeros（y_shape）），然后使用占位符初始化迭代器。

In my graph, I use tf_x and tf_y to do training, that is fine. But now I want to do referring, where I don't have label y. One workaround I made is to fake a y (like tf.zeros(y_shape)), then I use a placeholder to init the iterator.

x_placeholder = tf.placeholder(...)
y_placeholder = tf.placeholder(...)
ds = tf.data.Dataset.from_tensors((x_placeholder, y_placeholder))
ds_init_op = iterator.make_initializer(ds)
sess.run(ds_init_op, feed_dict={x_placeholder=x, y_placeholder=fake(y))})

我的问题是，有没有一种更清洁的方法？在推断时间内没有伪造 y ？

My question is, is there a cleaner way to do that? without fake a y during inferring time?

更新：

我做了一点实验，好像有一个数据集操作解压缩丢失了：

I experiment a little bit, looks like there is one dataset operation unzip missing:

import numpy as np
import tensorflow as tf


x_type = tf.float32
y_type = tf.float32
x_shape = tf.TensorShape([None, 128])
y_shape = tf.TensorShape([None, 10])
x_shape_nobatch = tf.TensorShape([128])
y_shape_nobatch = tf.TensorShape([10])

iterator_x = tf.data.Iterator.from_structure((x_type,), (x_shape,))
iterator_y = tf.data.Iterator.from_structure((y_type,), (y_shape,))


def gen1():
    for i in range(100):
        yield np.random.randn(128)
ds1 = tf.data.Dataset.from_generator(gen1, (x_type,), (x_shape_nobatch,))
ds1 = ds1.batch(5)
ds1_init_op = iterator_x.make_initializer(ds1)


def gen2():
    for i in range(80):
        yield np.random.randn(128), np.random.randn(10)
ds2 = tf.data.Dataset.from_generator(gen2, (x_type, y_type), (x_shape_nobatch, y_shape_nobatch))
ds2 = ds2.batch(10)

# my ds2 has two tensors in one element, now the problem is
# how can I unzip this dataset so that I can apply them to iterator_x and iterator_y?
# such as:
ds2_x, ds2_y = tf.data.Dataset.unzip(ds2)  #?? missing this unzip operation!
ds2_x_init_op = iterator_x.make_initializer(ds2_x)
ds2_y_init_op = iterator_y.make_initializer(ds2_y)


tf_x = iterator_x.get_next()
tf_y = iterator_y.get_next()

推荐答案

数据集API的目的是避免直接将值提供给会话（因为这会导致数据先流到客户端，然后流到设备）。

The purpose of datasets API is to avoid feeding the values directly to session (because that causes the data to flow first to the client, then to a device).

我看到的所有示例都使用数据集API还使用估算器API ，您可以在其中提供用于训练和推理的不同输入功能。

All examples I've seen that use datasets API also use estimator API, where you can provide different input functions for training and inference.

def train_dataset(data_dir):
  """Returns a tf.data.Dataset yielding (image, label) pairs for training."""
  data = input_data.read_data_sets(data_dir, one_hot=True).train
  return tf.data.Dataset.from_tensor_slices((data.images, data.labels))

def infer_dataset(data_dir):
  """Returns a tf.data.Dataset yielding images for inference."""
  data = input_data.read_data_sets(data_dir, one_hot=True).test
  return tf.data.Dataset.from_tensors((data.images,))

...

def train_input_fn():
  dataset = train_dataset(FLAGS.data_dir)
  dataset = dataset.shuffle(buffer_size=50000).batch(1024).repeat(10)
  (images, labels) = dataset.make_one_shot_iterator().get_next()
  return (images, labels)

mnist_classifier.train(input_fn=train_input_fn)

...

def infer_input_fn():
  return infer_dataset(FLAGS.data_dir).make_one_shot_iterator().get_next()

mnist_classifier.predict(input_fn=infer_input_fn)

这篇关于如何在训练和推理中使用tf.Dataset设计？的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！