def testDeserializeKerasModelTopologyAndWeightsFromBuffers(self):
"""Test loading of model and its weights from buffers."""
# Use separate tf.Graph and tf.Session contexts to prevent name collision.
with tf.Graph().as_default(), tf.Session():
tfjs_path = os.path.join(self._tmp_dir, 'model_for_test')
model1 = self._saveKerasModelForTest(tfjs_path)
model1_weight_values = model1.get_weights()
# Read the content of model.json into a BytesIO object.
with open(os.path.join(tfjs_path, 'model.json'), 'rb') as f:
json_buff = f.read()
weight_paths = sorted(glob.glob(os.path.join(tfjs_path, 'group*')))
weight_buffers = []
for path in weight_paths:
with open(path, 'rb') as f:
weight_buffers.append(f.read())
with tf.Graph().as_default(), tf.Session():
model2 = keras_tfjs_loader.deserialize_keras_model(
json_buff, weight_data=weight_buffers)
# Verify the equality of all the weight values.
model2_weight_values = model2.get_weights()
self.assertEqual(len(model1_weight_values),
len(model2_weight_values))
for model1_weight_value, model2_weight_value in zip(
model1_weight_values, model2_weight_values):
self.assertAllClose(model1_weight_value, model2_weight_value)
# The two model JSONs should match exactly.
self.assertEqual(model1.to_json(), model2.to_json())
def testDeserializeKerasModelTopologyAndWeightsFromFileObjects(self):
"""Test loading of model and its weights using file objects."""
# Use separate tf.Graph and tf.Session contexts to prevent name collision.
with tf.Graph().as_default(), tf.Session():
tfjs_path = os.path.join(self._tmp_dir, 'model_for_test')
model1 = self._saveKerasModelForTest(tfjs_path)
model1_weight_values = model1.get_weights()
# Read the content of model.json into a file object.
json_file = open(os.path.join(tfjs_path, 'model.json'), 'rb')
weight_paths = sorted(glob.glob(os.path.join(tfjs_path, 'group*')))
weight_files = [open(path, 'rb') for path in weight_paths]
with tf.Graph().as_default(), tf.Session():
model2 = keras_tfjs_loader.deserialize_keras_model(
json_file, weight_files)
# Verify the equality of all the weight values.
model2_weight_values = model2.get_weights()
self.assertEqual(len(model1_weight_values),
len(model2_weight_values))
for model1_weight_value, model2_weight_value in zip(
model1_weight_values, model2_weight_values):
self.assertAllClose(model1_weight_value, model2_weight_value)
# The two model JSONs should match exactly.
self.assertEqual(model1.to_json(), model2.to_json())
json_file.close()
for f in weight_files:
f.close()
def testDeserializeKerasModelTopologyOnlyFromJSONDict(self):
"""Test loading of model (only topology) from a JSON Dict."""
# Use separate tf.Graph and tf.Session contexts to prevent name collision.
with tf.Graph().as_default(), tf.Session():
tfjs_path = os.path.join(self._tmp_dir, 'model_for_test')
model1 = self._saveKerasModelForTest(tfjs_path)
# Read the content of model.json into a BytesIO
with open(os.path.join(tfjs_path, 'model.json'), 'rb') as f:
config_json = json.load(f)
with tf.Graph().as_default(), tf.Session():
model2 = keras_tfjs_loader.deserialize_keras_model(config_json)
# The two model JSONs should match exactly.
self.assertEqual(model1.to_json(), model2.to_json())
def testDeserializeKerasModelTopologyOnlyFromBytesIO(self):
"""Test loading of model (only topology) from a BytesIO object."""
# Use separate tf.Graph and tf.Session contexts to prevent name collision.
with tf.Graph().as_default(), tf.Session():
tfjs_path = os.path.join(self._tmp_dir, 'model_for_test')
model1 = self._saveKerasModelForTest(tfjs_path)
# Read the content of model.json into a BytesIO
buff = io.BytesIO()
buff_writer = io.BufferedWriter(buff)
with open(os.path.join(tfjs_path, 'model.json'), 'rb') as f:
buff_writer.write(f.read())
buff_writer.flush()
buff_writer.seek(0)
with tf.Graph().as_default(), tf.Session():
model2 = keras_tfjs_loader.deserialize_keras_model(buff)
# The two model JSONs should match exactly.
self.assertEqual(model1.to_json(), model2.to_json())
