日志记录和调试

此示例展示了如何为 auto-sklearn 提供自定义的日志配置。我们将拟合 2 个管道并在控制台上显示所有 INFO 级别的消息。即使您不提供 logging_configuration，autosklearn 也会在临时工作目录中创建一个日志文件。此目录可以通过 tmp_folder 来指定，如下例所示。

此示例还重点介绍了关于 auto-sklearn 内部目录结构的其他信息。

import pathlib

import sklearn.datasets
import sklearn.metrics
import sklearn.model_selection

import autosklearn.classification

数据加载

从 https://www.openml.org/d/3 加载 kr-vs-kp 数据集

X, y = data = sklearn.datasets.fetch_openml(data_id=3, return_X_y=True, as_frame=True)

X_train, X_test, y_train, y_test = sklearn.model_selection.train_test_split(
    X, y, random_state=1
)

创建日志配置

auto-sklearn 使用默认的 logging config。我们将instead 创建一个自定义配置，如下所示

logging_config = {
    "version": 1,
    "disable_existing_loggers": True,
    "formatters": {
        "custom": {
            # More format options are available in the official
            # `documentation <https://docs.pythonlang.cn/3/howto/logging-cookbook.html>`_
            "format": "%(asctime)s - %(name)s - %(levelname)s - %(message)s"
        }
    },
    # Any INFO level msg will be printed to the console
    "handlers": {
        "console": {
            "level": "INFO",
            "formatter": "custom",
            "class": "logging.StreamHandler",
            "stream": "ext://sys.stdout",
        },
    },
    "loggers": {
        "": {  # root logger
            "level": "DEBUG",
        },
        "Client-EnsembleBuilder": {
            "level": "DEBUG",
            "handlers": ["console"],
        },
    },
}

构建并拟合分类器

cls = autosklearn.classification.AutoSklearnClassifier(
    time_left_for_this_task=30,
    # Bellow two flags are provided to speed up calculations
    # Not recommended for a real implementation
    initial_configurations_via_metalearning=0,
    smac_scenario_args={"runcount_limit": 2},
    # Pass the config file we created
    logging_config=logging_config,
    # *auto-sklearn* generates temporal files under tmp_folder
    tmp_folder="./tmp_folder",
    # By default tmp_folder is deleted. We will preserve it
    # for debug purposes
    delete_tmp_folder_after_terminate=False,
)
cls.fit(X_train, y_train, X_test, y_test)

# *auto-sklearn* generates intermediate files which can be of interest
# Dask multiprocessing information. Useful on multi-core runs:
#   * tmp_folder/distributed.log
# The individual fitted estimators are written to disk on:
#   * tmp_folder/.auto-sklearn/runs
# SMAC output is stored in this directory.
# For more info, you can check the `SMAC documentation <https://github.com/automl/SMAC3>`_
#   * tmp_folder/smac3-output
# Auto-sklearn always outputs to this log file
# tmp_folder/AutoML*.log
for filename in pathlib.Path("./tmp_folder").glob("*"):
    print(filename)

/home/runner/work/auto-sklearn/auto-sklearn/autosklearn/data/target_validator.py:187: UserWarning: Fitting transformer with a pandas series which has the dtype category. Inverse transform may not be able preserve dtype when converting to np.ndarray
  warnings.warn(
2022-09-20 08:55:45,145 - Client-EnsembleBuilder - INFO - DummyFuture: ([{'Timestamp': Timestamp('2022-09-20 08:55:45.128931'), 'ensemble_optimization_score': 0.9886363636363636, 'ensemble_test_score': 0.9899874843554443}], 50)/SingleThreadedClient() Started Ensemble builder job at 2022.09.20-08.55.45 for iteration 0.
2022-09-20 08:55:48,103 - Client-EnsembleBuilder - INFO - DummyFuture: ([{'Timestamp': Timestamp('2022-09-20 08:55:48.084418'), 'ensemble_optimization_score': 0.9911616161616161, 'ensemble_test_score': 0.9887359198998749}], 50)/SingleThreadedClient() Started Ensemble builder job at 2022.09.20-08.55.48 for iteration 1.
tmp_folder/distributed.log
tmp_folder/smac3-output
tmp_folder/AutoML(1):fa2baf03-38c1-11ed-8830-892d16569fbe.log
tmp_folder/.auto-sklearn
tmp_folder/space.json