def build_from_ckpt(self, checkpoint_filename):
'''Restore from a saved model'''
from zoo.automl.model.base_pytorch_model import PytorchBaseModel
model = PytorchBaseModel(self.model_creator, self.optimizer_creator,
self.loss_creator)
model.restore(checkpoint_filename)
return model
def build_from_ckpt(self, checkpoint_filename):
'''Restore from a saved model'''
if self.backend == "pytorch":
from zoo.automl.model.base_pytorch_model import PytorchBaseModel
model = PytorchBaseModel(**self.params)
model.restore(checkpoint_filename)
return model
elif self.backend == "keras":
from zoo.automl.model.base_keras_model import KerasBaseModel
model = KerasBaseModel(**self.params)
model.restore(checkpoint_filename)
return model
def load(file_path):
'''
Load the TSPipeline to a folder
:param file_path: the folder location to load the pipeline
'''
import pickle
model_init_path = os.path.join(file_path, DEFAULT_MODEL_INIT_DIR)
model_path = os.path.join(file_path, DEFAULT_BEST_MODEL_DIR)
data_process_path = os.path.join(file_path, DEFAULT_DATA_PROCESS_DIR)
best_config_path = os.path.join(file_path, DEFAULT_BEST_CONFIG_DIR)
with open(model_init_path, "rb") as f:
model_init = pickle.load(f)
with open(data_process_path, "rb") as f:
data_process = pickle.load(f)
with open(best_config_path, "rb") as f:
best_config = pickle.load(f)
from zoo.automl.model.base_pytorch_model import PytorchBaseModel
best_model = PytorchBaseModel(**model_init)
best_model.restore(model_path)
return TSPipeline(best_model, best_config, **data_process)