def build_and_test_estimator(self, model_type):
"""Ensure that model trains and minimizes loss."""
model = wide_deep.build_estimator(self.temp_dir, model_type)
# Train for 1 step to initialize model and evaluate initial loss
model.train(
input_fn=lambda: wide_deep.input_fn(
TEST_CSV, num_epochs=1, shuffle=True, batch_size=1),
steps=1)
initial_results = model.evaluate(
input_fn=lambda: wide_deep.input_fn(
TEST_CSV, num_epochs=1, shuffle=False, batch_size=1))
# Train for 100 epochs at batch size 3 and evaluate final loss
model.train(
input_fn=lambda: wide_deep.input_fn(
TEST_CSV, num_epochs=100, shuffle=True, batch_size=3))
final_results = model.evaluate(
input_fn=lambda: wide_deep.input_fn(
TEST_CSV, num_epochs=1, shuffle=False, batch_size=1))
print('%s initial results:' % model_type, initial_results)
print('%s final results:' % model_type, final_results)
# Ensure loss has decreased, while accuracy and both AUCs have increased.
self.assertLess(final_results['loss'], initial_results['loss'])
self.assertGreater(final_results['auc'], initial_results['auc'])
self.assertGreater(final_results['auc_precision_recall'],
initial_results['auc_precision_recall'])
self.assertGreater(final_results['accuracy'], initial_results['accuracy'])
def build_and_test_estimator(self, model_type):
"""Ensure that model trains and minimizes loss."""
model = wide_deep.build_estimator(self.temp_dir, model_type)
# Train for 1 step to initialize model and evaluate initial loss
model.train(input_fn=wide_deep.input_fn(TEST_TRAINING_CSV,
num_epochs=1,
shuffle=True,
batch_size=1),
steps=1)
initial_results = model.evaluate(input_fn=wide_deep.input_fn(
TEST_TRAINING_CSV, num_epochs=1, shuffle=False, batch_size=1))
# Train for 40 steps at batch size 2 and evaluate final loss
model.train(input_fn=wide_deep.input_fn(TEST_TRAINING_CSV,
num_epochs=None,
shuffle=True,
batch_size=2),
steps=40)
final_results = model.evaluate(input_fn=wide_deep.input_fn(
TEST_TRAINING_CSV, num_epochs=1, shuffle=False, batch_size=1))
print('%s initial results:' % model_type, initial_results)
print('%s final results:' % model_type, final_results)
self.assertLess(final_results['loss'], initial_results['loss'])
def build_and_test_estimator(self, model_type):
"""Ensure that model trains and minimizes loss."""
model = wide_deep.build_estimator(self.temp_dir, model_type)
# Train for 1 step to initialize model and evaluate initial loss
model.train(
input_fn=lambda: wide_deep.input_fn(
TEST_CSV, num_epochs=1, shuffle=True, batch_size=1),
steps=1)
initial_results = model.evaluate(
input_fn=lambda: wide_deep.input_fn(
TEST_CSV, num_epochs=1, shuffle=False, batch_size=1))
# Train for 100 epochs at batch size 3 and evaluate final loss
model.train(
input_fn=lambda: wide_deep.input_fn(
TEST_CSV, num_epochs=100, shuffle=True, batch_size=3))
final_results = model.evaluate(
input_fn=lambda: wide_deep.input_fn(
TEST_CSV, num_epochs=1, shuffle=False, batch_size=1))
print('%s initial results:' % model_type, initial_results)
print('%s final results:' % model_type, final_results)
# Ensure loss has decreased, while accuracy and both AUCs have increased.
self.assertLess(final_results['loss'], initial_results['loss'])
self.assertGreater(final_results['auc'], initial_results['auc'])
self.assertGreater(final_results['auc_precision_recall'],
initial_results['auc_precision_recall'])
self.assertGreater(final_results['accuracy'], initial_results['accuracy'])
def test_input_fn(self):
features, labels = wide_deep.input_fn(self.input_csv, 1, False, 1)
with tf.Session() as sess:
features, labels = sess.run((features, labels))
# Compare the two features dictionaries.
for key in TEST_INPUT_VALUES:
self.assertTrue(key in features)
self.assertEqual(len(features[key]), 1)
feature_value = features[key][0]
# Convert from bytes to string for Python 3.
if isinstance(feature_value, bytes):
feature_value = feature_value.decode()
self.assertEqual(TEST_INPUT_VALUES[key], feature_value)
self.assertFalse(labels)
def test_input_fn(self):
dataset = wide_deep.input_fn(self.input_csv, 1, False, 1)
features, labels = dataset.make_one_shot_iterator().get_next()
with tf.Session() as sess:
features, labels = sess.run((features, labels))
# Compare the two features dictionaries.
for key in TEST_INPUT_VALUES:
self.assertTrue(key in features)
self.assertEqual(len(features[key]), 1)
feature_value = features[key][0]
# Convert from bytes to string for Python 3.
if isinstance(feature_value, bytes):
feature_value = feature_value.decode()
self.assertEqual(TEST_INPUT_VALUES[key], feature_value)
self.assertFalse(labels)
def input_fn():
return wide_deep.input_fn(TEST_CSV,
num_epochs=num_epochs,
shuffle=shuffle,
batch_size=batch_size)
