Minimized TensorFlow regression model

tf_regression.py

import os.path
import time

import numpy as np
import tensorflow as tf

from tensorflow.python.training import saver as saver_lib
from tensorflow.python.training import summary_io

# Basic model parameters as external flags.
flags = tf.app.flags
FLAGS = flags.FLAGS
flags.DEFINE_float('learning_rate', 0.01, 'Initial learning rate.')
flags.DEFINE_integer('num_steps', 10000, 'Number of steps to run trainer.')
flags.DEFINE_integer('hidden_layers', 5, 'Number of hidden layers.')
flags.DEFINE_integer('hidden_layers_size', 64, 'Number of units in hidden layers.')
flags.DEFINE_integer('batch_size', 10, 'Batch size.  ')


def input_pipeline(batch_size):
  label = tf.constant([[10.1], [1.3], [12.5], [372.0], [443.0], [0.3], [23.45], [1.42], [0.0], [56.7]], dtype=tf.float32)
  example = tf.constant([[11.0], [11.0], [13.0], [370.0], [422.0], [125.0], [25.0], [15.0], [100.0], [55.0]], dtype=tf.float32)

  example, label = tf.train.slice_input_producer(
      [example, label],
      shuffle=True,
      capacity=30+3*batch_size)

  example_batch, label_batch = tf.train.batch(
      [example, label],
      batch_size=batch_size,
      capacity=30+3*batch_size)
  return example_batch, label_batch

def input_fn():
  return input_pipeline(FLAGS.batch_size)

def model_fn(features, target):
  # features = tf.Print(features, [features], message="This is features: ")
  # target = tf.Print(target, [target], message="This is target: ")
  # return {}, tf.constant(0), tf.group(features, target)
  out = tf.contrib.layers.stack(features,
                                tf.contrib.layers.fully_connected,
                                [FLAGS.hidden_layers_size] * FLAGS.hidden_layers,
                                activation_fn=tf.nn.relu)
  prediction = tf.contrib.layers.fully_connected(
      out,
      num_outputs=1,
      activation_fn=None)

  loss = tf.nn.l2_loss(prediction - target)
  train_op = tf.contrib.layers.optimize_loss(
      loss,
      tf.contrib.framework.get_global_step(),
      optimizer='Adagrad',
      learning_rate=FLAGS.learning_rate)

  rmse, update_op = tf.contrib.metrics.streaming_root_mean_squared_error(prediction, target)
  tf.scalar_summary("rmse", rmse)
  return prediction, loss, tf.group(train_op, update_op)

def create_eval_metrics():
  return {'rmse': tf.contrib.metrics.streaming_root_mean_squared_error}

def main(_):
  model_dir = time.strftime("/tmp/tf_regression/%m-%d-%H-%M-%S", time.gmtime())
  estimator = tf.contrib.learn.Estimator(model_fn=model_fn, model_dir=model_dir)
  estimator.fit(
      input_fn=input_fn,
      steps=FLAGS.num_steps,
      monitors=[])

if __name__ == '__main__':
  tf.app.run()