def convert_model_saved(flags, folder, mode, weights_name='best_weights'):
"""Convert model to streaming and non streaming SavedModel.
Args:
flags: model and data settings
folder: folder where converted model will be saved
mode: inference mode
weights_name: file name with model weights
"""
tf.reset_default_graph()
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
sess = tf.Session(config=config)
tf.keras.backend.set_session(sess)
tf.keras.backend.set_learning_phase(0)
flags.batch_size = 1 # set batch size for inference
model = models.MODELS[flags.model_name](flags)
weights_path = os.path.join(flags.train_dir, weights_name)
model.load_weights(weights_path).expect_partial()
path_model = os.path.join(flags.train_dir, folder)
if not os.path.exists(path_model):
os.makedirs(path_model)
try:
# convert trained model to SavedModel
utils.model_to_saved(model, flags, path_model, mode)
except IOError as e:
logging.warning('FAILED to write file: %s', e)
except (ValueError, AttributeError, RuntimeError, TypeError,
AssertionError) as e:
logging.warning('WARNING: failed to convert to SavedModel: %s', e)
def test_model_to_saved(self, model_name='dnn'):
"""SavedModel supports both stateless and stateful graphs."""
params = model_params.HOTWORD_MODEL_PARAMS[model_name]
params = model_flags.update_flags(params)
# create model
model = models.MODELS[params.model_name](params)
utils.model_to_saved(model, params, FLAGS.test_tmpdir)
def test_model_to_saved(self):
"""SavedModel supports both stateless and stateful graphs."""
utils.model_to_saved(self.model, self.flags, FLAGS.test_tmpdir)
