def __init__(
self,
tuner: Optional[SmacPipelineTuner] = None, # 抽象化的优化的全过程
hdl_constructor: Optional[HDL_Constructor] = None, # 用户自定义初始超参
resource_manager: Union[ResourceManager, str] = None,
ensemble_builder: Union[StackEnsembleBuilder, None, bool,
int] = None):
if ensemble_builder is None:
print("info: 使用默认的stack集成学习器")
ensemble_builder = StackEnsembleBuilder()
elif ensemble_builder == False:
print("info: 不使用集成学习")
else:
ensemble_builder = StackEnsembleBuilder(set_model=ensemble_builder)
self.ensemble_builder = ensemble_builder
if not tuner:
tuner = SmacPipelineTuner()
self.tuner: SmacPipelineTuner = tuner
if not hdl_constructor:
hdl_constructor = HDL_Constructor()
if isinstance(hdl_constructor, dict):
# todo: 使用用户自定义超参描述语言
print("使用用户自定义超参描述语言")
self.hdl_constructor = hdl_constructor
self.random_state = tuner.random_state
if isinstance(resource_manager, str):
resource_manager = ResourceManager(
resource_manager) # todo : 识别不同协议的文件系统，例如hdfs
elif resource_manager is None:
resource_manager = ResourceManager()
self.resource_manager = resource_manager
self.estimator = None
import pandas as pd
from sklearn.model_selection import ShuffleSplit
from autopipeline.estimator.base import AutoPipelineEstimator
from autopipeline.tuner.smac_tuner import SmacPipelineTuner
df = pd.read_csv("../examples/classification/train_classification.csv")
ss = ShuffleSplit(n_splits=1, random_state=0, test_size=0.25)
train_ix, test_ix = next(ss.split(df))
df_train = df.iloc[train_ix, :]
df_test = df.iloc[test_ix, :]
tuner = SmacPipelineTuner(
random_state=42,
initial_runs=5,
runcount_limit=12,
)
auto_pipeline = AutoPipelineEstimator(tuner)
column_descriptions = {
"id": "PassengerId",
"target": "Survived",
"ignore": "Name"
}
auto_pipeline.fit(X=df_train,
X_test=df_test,
column_descriptions=column_descriptions,
n_jobs=1)
estimators = []
for i in range(100):
hdl2phps = SmacHDL2PHPS()
hdl2phps.set_task(constants.binary_classification_task)
phps = hdl2phps(HDL)
# print(phps)
php = phps.sample_configuration()
# print(php)
php2dhp = SmacPHP2DHP()
dhp = php2dhp(php)
# i+=1
estimators.append(list(dhp["estimator"].keys())[0])
pprint(dhp)
break
# print(Counter(estimators))
tuner = SmacPipelineTuner()
tuner.set_task(constants.binary_classification_task)
preprocessor = tuner.create_preprocessor(dhp)
estimators = tuner.create_estimator(dhp)
pipeline = concat_pipeline(preprocessor, estimators)
print(pipeline)
df = pd.read_csv("../examples/classification/train_classification.csv")
y = df.pop("Survived").values
df = df.loc[:, [
"Pclass", "Name", "Sex", "Age", "SibSp", "Ticket", "Fare", "Cabin",
"Embarked"
]]
df_train, df_test, y_train, y_test = train_test_split(df,
y,
test_size=0.2,
random_state=10)