文档中心

XGBoost 1.3.1代码规范

基于XGBoost 1.3.1框架的结构化数据的回归问题。
如下所示是其超参搜索任务中一个超参数组合的训练代码，代码会通过argparse模块接受在平台中填写的信息，请保持一致。另外该框架支持发布保存模型为pickle和joblib格式，并且在发布至模型仓库时需要选择相应的模型文件。

xgboost1.3.1_autosearch.py示例代码

# -*- coding:utf-8 -*- """ xgboost train demo """ import xgboost as xgb from sklearn.model_selection

 import train_test_split from sklearn import datasets from sklearn.metrics import mean_squared_error

 import numpy as np import os import time import argparse from rudder_autosearch.sdk.amaas_tools import 

AMaasTools def parse_arg(): """parse arguments""" parser = argparse.ArgumentParser

(description='xgboost boston Example') parser.add_argument('--train_dir', type=str, 

default='./train_data', help='input data dir for training (default: ./train_data)')

 parser.add_argument('--test_dir', type=str, default='./test_data', help='input data dir 

for test (default: ./test_data)') parser.add_argument('--output_dir', type=str, default='

./output', help='output dir for auto_search job (default: ./output)') parser.add_argument(

'--job_id', type=str, default="job-1234", help='auto_search job id (default: "job-1234")')

 parser.add_argument('--trial_id', type=str, default="0-0", help='auto_search id of a single 

trial (default: "0-0")') parser.add_argument('--metric', type=str, default="mse", help='evaluation 

metric of the model') parser.add_argument('--data_sampling_scale', type=float, default=1.0, help=

'sampling ratio of the dataset for auto_search (default: 1.0)') parser.add_argument('--max_depth',

 type=int, default=6, help='maximum depth of the tree (default: 6)') parser.add_argument('--gamma',

 type=float, default=0.1, help='minimum loss reduction required for further splitting (default: 0.1)')

 parser.add_argument('--eta', type=float, default=0.1, help='learning rate (default: 0.1)') 

parser.add_argument('--num_round', type=int, default=10, help='number of trees (default: 10)')

 args = parser.parse_args() args.output_dir = os.path.join(args.output_dir, args.job_id, args.trial_id) 

if not os.path.exists(args.output_dir): os.makedirs(args.output_dir) print("job_id: {}, trial_id:

 {}".format(args.job_id, args.trial_id)) return args def load_data(data_sampling_scale): "

"" load data """ boston = datasets.load_boston() X, Y = boston.data, boston.target 

# 切分，测试训练2,8分 x_train, x_test, y_train, y_test = train_test_split(X, Y, test_size=0.2, 

random_state=0) train_data = np.concatenate([x_train, y_train.reshape([-1, 1])], axis=1) np.random.seed(0)

 np.random.shuffle(train_data) train_data = train_data[0:int(data_sampling_scale * len(train_data))] x_train, 

y_train = train_data[:, 0:-1], train_data[:, -1] return (x_train, x_test), (y_train, y_test) def save_model

(model, output_dir): """ save model with pickle format """ import pickle with open(output_dir 

+ '/clf.pickle', 'wb') as f: pickle.dump(model, f) def save_model_joblib(model, output_dir): 

""" save model with joblib format """ import joblib
    joblib.dump(model, output_dir + '/clf.pkl') def evaluate(model, x_test, y_test): "

""evaluate""" # 回归mean_squared_error指标 deval = xgb.DMatrix(x_test) predict = model.predict(deval)

 mse = mean_squared_error(y_test, predict) print("mean_squared_error: %f" % mse) return mse def report_

final(args, metric): """report_final_result""" # 结果上报sdk amaas_tools = AMaasTools(args.job_id, 

args.trial_id) metric_dict = {args.metric: metric} for i in range(3): flag, ret_msg = amaas_tools.report

_final_result(metric=metric_dict, export_model_path=args.output_dir, checkpoint_path="") print("End

 Report, metric:{}, ret_msg:{}".format(metric, ret_msg)) if flag: break time.sleep(1) assert flag, 

"Report final result to manager failed! Please check whether manager'address or manager'status "

 \ "is ok! " def main(): """ main """ # 获取参数 args = parse_arg() # 加载数据集 (x_train, x_test),

 (y_train, y_test) = load_data(args.data_sampling_scale) dtrain = xgb.DMatrix(x_train, label=y_train) 

# 模型参数定义 param = {"gamma": args.gamma, 'max_depth': args.max_depth, 'eta': args.eta, 'objective'

: 'reg:squarederror'} # 模型训练 model = xgb.train(param, dtrain, args.num_round) 

# 模型保存 save_model_joblib(model, args.output_dir) # 模型评估 mse = evaluate(model, x_test, y_test) 

# 上报结果 report_final(args, metric=mse) if __name__ == '__main__': main()

示例代码对应的yaml配置如下，请保持格式一致

cmaes_search_demo.yml示例内容

#搜索算法参数
search_strategy:
  algo: CMAES_SEARCH #搜索策略：进化-cmaes算法
  params:
    population_num: 8 #种群个体数量 ｜ [1,10] int类型
    round: 10 #迭代轮数     ｜[5,50] int类型
    step_size: 1.0 # 学习步长 ｜（0，10] float类型

#单次训练时数据的采样比例，单位%
data_sampling_scale: 100  #｜（0，100] int类型

#评价指标参数
metrics:
  name: mse #评价指标 ｜ 任意字符串 str类型
  goal: MINIMIZE #最大值/最小值 ｜ str类型   MAXIMIZE or MINIMIZE   必须为这两个之一（也即支持大写）
  expected_value: 10 #早停标准值，评价指标超过该值则结束整个超参搜索，单位% ｜无限制 int类型

#搜索参数空间
search_space:
  max_depth:
    htype: randint
    value: [3, 10]
  num_round:
    htype: randint
    value: [1, 8]
  gamma:
    htype: uniform
    value: [0.1, 1]
  eta:
    htype: loguniform
    value: [0.01, 1]