ProblemInfo 类
有关正在处理的问题的元数据的容器对象。
提供可用于预测未来管道的更多信息。当数据无法直接用于检查时,
该信息对于预测管道成本非常重要。
- 继承
-
builtins.objectProblemInfo
构造函数
ProblemInfo(dataset_samples=0, dataset_features=0, dataset_classes=0, dataset_num_categorical=0, dataset_categoricals=None, pipeline_categoricals=None, dataset_y_std=0, dataset_uid=None, runtime_constraints=None, num_threshold_buffers=3, num_recommendations=1, num_threads=1, is_sparse=False, pipeline_profile='none', constraint_mode=1, cost_mode=1, task='classification', training_percent=None, subsampling=False, metric='AUC_macro', model_names_whitelisted=None, model_names_blacklisted=None, kernel='linear', subsampling_treatment='linear', subsampling_schedule='hyperband_clip', cost_mode_param=None, iteration_timeout_mode=0, iteration_timeout_param=None, feature_column_names=None, label_column_name=None, weight_column_name=None, cv_split_column_names=None, enable_streaming=None, timeseries_param_dict=None, gpu_training_param_dict=None, max_time=None)
参数
- dataset_samples
- dataset_features
- dataset_classes
- dataset_num_categorical
- dataset_categoricals
- pipeline_categoricals
- dataset_y_std
- dataset_uid
- runtime_constraints
- num_threshold_buffers
- num_recommendations
- num_threads
- is_sparse
- pipeline_profile
- constraint_mode
- cost_mode
- task
- training_percent
- subsampling
- metric
- model_names_whitelisted
- model_names_blacklisted
- kernel
- subsampling_treatment
- subsampling_schedule
- cost_mode_param
- iteration_timeout_mode
- iteration_timeout_param
- feature_column_names
- label_column_name
- weight_column_name
- cv_split_column_names
- enable_streaming
- timeseries_param_dict
- gpu_training_param_dict
- max_time
方法
| add_pipeline_profile |
添加管道配置文件。 其中管道配置文件是要应用的管道掩码。 |
| clean_attrs |
清理所有不需要的特性的 ProblemInfo |
| clean_dataset_info |
清理所有可能敏感的用户数据的 ProblemInfo。 |
| done |
如果客户端应暂停训练,则返回 true。 |
| from_dict |
创建 ProblemInfo 对象。 |
| get_default_parameters_and_values |
检查 ProblemInfo 构造函数的参数,并返回分配了默认值的所有参数。 |
| get_time_constraint |
获取置于试验上的时间约束。 |
| handle_server_code |
在客户端进行调用以解释服务器代码。 |
| set_cost_mode |
设置问题的成本模式。 |
| set_pipeline_profile |
设置管道配置文件。 其中管道配置文件是要应用的管道掩码。 |
| set_runtime_constraint |
设置运行时约束。 |
| set_start_time |
设置开始时间。 |
| set_time_constraint |
获取置于试验上的时间约束。 |
| to_dict |
将当前 ProblemInfo 对象转换为字典。 |
| update_pipeline_profile |
更新管道配置文件。 其中管道配置文件是要应用的管道掩码。 |
| update_time |
更新当前时间。 |
add_pipeline_profile
添加管道配置文件。
其中管道配置文件是要应用的管道掩码。
add_pipeline_profile(pipeline_profile)
参数
- pipeline_profile
clean_attrs
清理所有不需要的特性的 ProblemInfo
clean_attrs(useful_attrs: List[str]) -> azureml.automl.runtime.shared.problem_info.ProblemInfo
参数
- useful_attrs
clean_dataset_info
清理所有可能敏感的用户数据的 ProblemInfo。
clean_dataset_info()
done
如果客户端应暂停训练,则返回 true。
done(final_fit_expected_time=None)
参数
- final_fit_expected_time
from_dict
创建 ProblemInfo 对象。
static from_dict(d)
参数
- d
数据集特性的字典
返回
ProblemInfo 对象
get_default_parameters_and_values
检查 ProblemInfo 构造函数的参数,并返回分配了默认值的所有参数。
get_default_parameters_and_values()
参数
- cls
get_time_constraint
获取置于试验上的时间约束。
get_time_constraint()
返回
时间约束
handle_server_code
在客户端进行调用以解释服务器代码。
handle_server_code(status_code)
参数
- status_code
从服务器发送的代码,例如,用于增加时间约束的代码。
- runner
还要更新的运行程序对象(如有必要)。
set_cost_mode
设置问题的成本模式。
set_cost_mode()
set_pipeline_profile
设置管道配置文件。
其中管道配置文件是要应用的管道掩码。
set_pipeline_profile(pipeline_profile)
参数
- pipeline_profile
set_runtime_constraint
设置运行时约束。
set_runtime_constraint(new_constraint)
参数
- new_constraint
要设置的约束的字典。 约束的键在 default_resource_limits resource_limits 中,如果约束键在字典中不存在,则它将保持不变。
set_start_time
设置开始时间。
set_start_time(t)
参数
- t
set_time_constraint
获取置于试验上的时间约束。
set_time_constraint(new_constraint)
参数
- new_constraint
int,时间约束(秒)
to_dict
将当前 ProblemInfo 对象转换为字典。
to_dict()
返回
数据集特性的字典
update_pipeline_profile
更新管道配置文件。
其中管道配置文件是要应用的管道掩码。
update_pipeline_profile(pipeline_profile)
参数
- pipeline_profile
update_time
更新当前时间。
update_time(t, training_times=0, predict_times=0)
参数
- t
当前时间,按 time.time() 进行度量
- training_times
当使用缓存的结果来考虑所跳过的时间时使用的可选参数
- predict_times
用于忽略预测管道所用时间的可选参数
反馈
提交和查看相关反馈