def load_from_checkpoint(
checkpoint: Checkpoint, ) -> Tuple[xgb.Booster, Optional[Preprocessor]]:
checkpoint_path = checkpoint.to_directory()
xgb_model = xgb.Booster()
xgb_model.load_model(os.path.join(checkpoint_path, MODEL_KEY))
preprocessor_path = os.path.join(checkpoint_path, PREPROCESSOR_KEY)
if os.path.exists(preprocessor_path):
with open(preprocessor_path, "rb") as f:
preprocessor = cpickle.load(f)
else:
preprocessor = None
return xgb_model, preprocessor
def load_from_checkpoint(
checkpoint: Checkpoint,
) -> Tuple[RandomForestClassifier, Optional[Preprocessor]]:
path = checkpoint.to_directory()
estimator_path = os.path.join(path, MODEL_KEY)
with open(estimator_path, "rb") as f:
estimator = cpickle.load(f)
preprocessor_path = os.path.join(path, PREPROCESSOR_KEY)
if os.path.exists(preprocessor_path):
with open(preprocessor_path, "rb") as f:
preprocessor = cpickle.load(f)
else:
preprocessor = None
return estimator, preprocessor
def from_checkpoint(cls, checkpoint: Checkpoint) -> "LightGBMPredictor":
"""Instantiate the predictor from a Checkpoint.
The checkpoint is expected to be a result of ``LightGBMTrainer``.
Args:
checkpoint (Checkpoint): The checkpoint to load the model and
preprocessor from. It is expected to be from the result of a
``LightGBMTrainer`` run.
"""
path = checkpoint.to_directory()
bst = lightgbm.Booster(model_file=os.path.join(path, MODEL_KEY))
preprocessor_path = os.path.join(path, PREPROCESSOR_KEY)
if os.path.exists(preprocessor_path):
with open(preprocessor_path, "rb") as f:
preprocessor = cpickle.load(f)
else:
preprocessor = None
shutil.rmtree(path)
return LightGBMPredictor(model=bst, preprocessor=preprocessor)
