文档中心

Sklearn 0.23.2代码规范

基于Sklearn 0.23.2框架的结构化数据的多分类问题，训练数据集sklearn_train_data.zip点击这里下载。
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。

sklearn0.23.2_autosearch.py示例代码

# -*- coding:utf-8 -*- """ sklearn train demo """ import os import argparse import

 time from sklearn.model_selection import train_test_split from sklearn.metrics import

 f1_score from sklearn import svm import pandas as pd import numpy as np from rudder_

autosearch.sdk.amaas_tools import AMaasTools def parse_arg(): """parse arguments"""

 parser = argparse.ArgumentParser(description='Sklearn iris 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') parser.add_argument('--trial_id', type=str, default="0-0", 

help='auto_search id of a single trial') parser.add_argument('--metric', type=str,

 default="f1_score", 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('--kernel', type=str,

 default='linear', help='kernel function (default: "linear")') parser.add_argument

('--C', type=float, default=1, help='penalty term (default: 1)') parser.add_argument

('--gamma', type=float, default=0.5, help='parameter of the kernel (default: 0.5)')

 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(train_dir, data_sampling_scale): """ load data """ # 共150条数据，

训练120条，测试30条，进行2,8分进行模型训练 # 每条数据类型为 x{nbarray} [6.4, 3.1, 5.5, 1.8] 

# 上传的数据储存在./train_data和./test_data中 inputdata = pd.read_csv(train_dir + "/iris.csv") 

target = inputdata["Species"] inputdata = inputdata.drop(columns=["Species"]) # 切分，测试训练2,8分 

x_train, x_test, y_train, y_test = train_test_split(inputdata, target, test_size=0.2, random_state=0)

 train_data = np.concatenate([x_train, y_train.ravel().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 """ try: import joblib except: from sklearn.externals import joblib
    joblib.dump(model, output_dir + '/clf.pkl') def evaluate(model, x_test, y_test): """evaluate"""

 # 多分类f1_score指标 predict = model.predict(x_test) f1 = f1_score(y_test, predict, average="micro")

 print("f1_score: %f" % f1) return f1 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.train_dir, args.data_sampling_scale) # 模型定义 model =

 svm.SVC(C=args.C, kernel=args.kernel, gamma=args.gamma) # 模型训练 model.fit(x_train, y_train)

 # 模型保存 save_model(model, args.output_dir) # 模型评估 f1 = evaluate(model, x_test, y_test) 

# 上报结果 report_final(args, metric=f1) if __name__ == '__main__': main()

示例代码对应的yaml配置如下，请保持格式一致

random_search_demo.yml示例内容

#搜索算法参数
search_strategy:
  algo: RANDOM_SEARCH #搜索策略：随机搜索

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

#最大搜索次数
max_trial_num: 10 # ｜>0 int类型

#评价指标参数
metrics:
  name: f1_score #评价指标 ｜ 任意字符串 str类型
  goal: MAXIMIZE #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: 100 #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:
  kernel: #核函数
    htype: choice
    value: ["linear", "rbf"]
  C: #惩罚项
    htype: loguniform
    value: [0.001, 1000]
  gamma: #核函数参数
    htype: loguniform
    value: [0.0001, 1]