def _fe_compile(model: Model, optimizer_fn: Union[str, Scheduler, Callable,
None],
weight: Union[str, None], name: str, framework: str) -> Model:
"""A function to bundle models with their optimizers.
Args:
model: The model to be bundled.
optimizer_fn: The optimizer to be associated with the given `model`.
weight: A path to weights to be associated with the `model`.
name: The name of the model.
framework: Which backend framework should be associated with this model (either 'tf' or 'torch').
Returns:
The `model` combined with its optimizer, weights, and name. Models will also have an 'fe_compiled' flag to
indicate that they were built via this function.
"""
if isinstance(optimizer_fn, EpochScheduler):
for epoch, optimizer_def in optimizer_fn.epoch_dict.items():
optimizer_fn.epoch_dict[epoch] = _build_optimizer(
optimizer_def, model, framework)
model.current_optimizer = optimizer_fn.get_current_value(1)
elif isinstance(optimizer_fn, RepeatScheduler):
for idx, optimizer_def in enumerate(optimizer_fn.repeat_list):
optimizer_fn.repeat_list[idx] = _build_optimizer(
optimizer_def, model, framework)
model.current_optimizer = optimizer_fn.get_current_value(1)
else:
optimizer_fn = _build_optimizer(optimizer_fn, model, framework)
model.current_optimizer = optimizer_fn
model.optimizer = optimizer_fn
model.fe_compiled = True
if weight:
if weight.startswith(GOOGLE_DRIVE_URL):
tmp_dir = tempfile.mkdtemp()
file_name = gdown.download(weight, quiet=False)
os.rename(os.path.join('./', file_name),
os.path.join(tmp_dir, file_name))
weight = gdown.download(weight,
os.path.join(tmp_dir, file_name),
quiet=False)
load_model(model, weight)
model.model_name = name
return model
def _load_files(self) -> None:
"""Restore from files.
"""
system_path = os.path.join(self.directory, self.system_file)
self.system.load_state(json_path=system_path)
for model in self.system.network.models:
if isinstance(model, tf.keras.Model):
framework = "tf"
elif isinstance(model, torch.nn.Module):
framework = "torch"
else:
raise ValueError("Unknown model type {}".format(type(model)))
weights_path = os.path.join(
self.directory,
"{}.{}".format(model.model_name,
self.model_extension[framework]))
load_model(model, weights_path=weights_path, load_optimizer=True)
def on_end(self, data: Data) -> None:
if self.load_best_final and self.model_path:
print(
"FastEstimator-BestModelSaver: Restoring model from {}".format(
self.model_path))
load_model(self.model, self.model_path)