Bases: ModelLoader
Interface implementation to load a keras model
Constructor for KerasModelLoader
:param model_custom_objects: Optional custom objects to load the model.
In Keras it is possible to pass custom objects during model loading.
:param compile_model: Flag if the model should be compiled after loading
Source code in niceml/dlframeworks/keras/kerasmodelloader.py
| def __init__(
self,
model_custom_objects: Optional[ModelCustomLoadObjects] = None,
compile_model: bool = False,
):
"""
Constructor for KerasModelLoader
:param model_custom_objects: Optional custom objects to load the model.
In Keras it is possible to pass custom objects during model loading.
:param compile_model: Flag if the model should be compiled after loading
"""
self.compile_model = compile_model
self.model_custom_objects: ModelCustomLoadObjects = (
model_custom_objects or ModelCustomLoadObjects()
)
|
Functions
__call__
__call__(model_path, file_system=None)
Loads the model at the given path
Source code in niceml/dlframeworks/keras/kerasmodelloader.py
| def __call__(
self,
model_path: str,
file_system: Optional[AbstractFileSystem] = None,
) -> Any:
"""Loads the model at the given path"""
file_system = file_system or LocalFileSystem()
with TemporaryDirectory() as tmp_dir:
tmp_model_path = join(tmp_dir, "model.hdf5")
with open(tmp_model_path, "wb") as tmp_model_file, file_system.open(
model_path, "rb"
) as model_file:
tmp_model_file.write(model_file.read())
model = load_model(
tmp_model_path, self.model_custom_objects(), compile=self.compile_model
)
return model
|