文档首页/ AI开发平台ModelArts/ 推理部署（即将下线）/ 推理规范说明/ 自定义脚本代码示例/ Pyspark
更新时间：2024-10-30 GMT+08:00
查看PDF
分享

Pyspark

训练并保存模型

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
 
from pyspark.ml import Pipeline, PipelineModel
from pyspark.ml.linalg import Vectors
from pyspark.ml.classification import LogisticRegression

# 创建训练数据，此处通过tuples创建
# Prepare training data from a list of (label, features) tuples.
training = spark.createDataFrame([
    (1.0, Vectors.dense([0.0, 1.1, 0.1])),
    (0.0, Vectors.dense([2.0, 1.0, -1.0])),
    (0.0, Vectors.dense([2.0, 1.3, 1.0])),
    (1.0, Vectors.dense([0.0, 1.2, -0.5]))], ["label", "features"])

# 创建训练实例，此处使用逻辑回归算法进行训练
# Create a LogisticRegression instance. This instance is an Estimator.
lr = LogisticRegression(maxIter=10, regParam=0.01)

# 训练逻辑回归模型
# Learn a LogisticRegression model. This uses the parameters stored in lr.
model = lr.fit(training)

# 保存模型到本地目录
# Save model to local path.
model.save("/tmp/spark_model")

保存完模型后，需要上传到OBS目录才能发布。发布时需要带上config.json配置和推理代码customize_service.py。config.json编写请参考模型配置文件编写说明，推理代码请参考推理代码。

推理代码

在模型代码推理文件customize_service.py中，需要添加一个子类，该子类继承对应模型类型的父类，各模型类型的父类名称和导入语句如请参考表1。

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
 
# coding:utf-8
import collections
import json
import traceback

import model_service.log as log
from model_service.spark_model_service import SparkServingBaseService
from pyspark.ml.classification import LogisticRegression

logger = log.getLogger(__name__)


class UserService(SparkServingBaseService):
    # 数据预处理
    def _preprocess(self, data):
        logger.info("Begin to handle data from user data...")
        # 读取数据
        req_json = json.loads(data, object_pairs_hook=collections.OrderedDict)
        try:
            # 将数据转换成spark dataframe格式
            predict_spdf = self.spark.createDataFrame(pd.DataFrame(req_json["data"]["req_data"]))
        except Exception as e:
            logger.error("check your request data does meet the requirements ?")
            logger.error(traceback.format_exc())
            raise Exception("check your request data does meet the requirements ?")
        return predict_spdf

    # 模型推理
    def _inference(self, data):
        try:
            # 加载模型文件
            predict_model = LogisticRegression.load(self.model_path)
            # 对数据进行推理
            prediction_result = predict_model.transform(data)
	except Exception as e:
            logger.error(traceback.format_exc())
            raise Exception("Unable to load model and do dataframe transformation.")
        return prediction_result

    # 数据后处理
    def _postprocess(self, pre_data):
        logger.info("Get new data to respond...")
        predict_str = pre_data.toPandas().to_json(orient='records')
        predict_result = json.loads(predict_str)
        return predict_result