Tensorflow Custom Estimator

tf_estimator_custom.py

import numpy as np
import tensorflow as tf

# 커스텀 Estimator 모델(func)
def model_fn(features, labels, mode):
	W = tf.get_variable('W', [1], dtype=tf.float64) # tf.get_variable(): Gets an existing variable with parameter or create a new one.
	b = tf.get_variable('b', [1], dtype=tf.float64)
	y = W * features['x'] + b
	
	# loss(오차) 서브 그래프
	loss = tf.reduce_sum(tf.square(y - labels)) # labels: 지도 해답?
	
	# 훈련 서브 그래프
	global_step = tf.train.get_global_step()
	optimizer = tf.train.GradientDescentOptimizer(0.01)
	train = tf.group(optimizer.minimize(loss), tf.assign_add(global_step, 1)) # tf.group(): Create an op that groups multiple operations.
	
	return tf.estimator.EstimatorSpec(
		mode = mode,
		predictions = y,
		loss = loss,
		train_op = train)

# 커스텀 Estimator
estimator = tf.estimator.Estimator(model_fn = model_fn)

# 데이터 세트
x_train = np.array([1.0, 2.0, 3.0, 4.0])
y_train = np.array([0.0, -1.0, -2.0, -3.0])
x_eval = np.array([2.0, 5.0, 8.0, 1.0])
y_eval = np.array([-1.01, -4.1, -7.0, 0.0])

input_fn = tf.estimator.inputs.numpy_input_fn(
	{'x': x_train}, y_train, batch_size=4, num_epochs=None, shuffle=True)
train_input_fn = tf.estimator.inputs.numpy_input_fn(
	{'x': x_train}, y_train, batch_size=4, num_epochs=1000, shuffle=False)
eval_input_fn = tf.estimator.inputs.numpy_input_fn(
	{'x': x_eval}, y_eval, batch_size=4, num_epochs=1000, shuffle=False)

# 훈련
estimator.train(input_fn=input_fn, steps=1000)

# 결과
train_metrics = estimator.evaluate(input_fn=train_input_fn)
eval_metrics = estimator.evaluate(input_fn=eval_input_fn)
print('train metrics: %r' % train_metrics)
print('eval metrics: %r' % eval_metrics)

Thank you so much for sharing this code snippet!!!! I've been looking for such an example!