iailab-plat.git

houzhongjian

2024-12-26 b482efaf698d5fd3920ff63ccde1c819f9eb34ba

提交 \| 用户 \| 时间
7fd198	1	package com.iailab.module.model.mdk.predict.impl;
潘	2
6957a3	3	import com.alibaba.fastjson.JSON;
7fd198	4	import com.alibaba.fastjson.JSONArray;
潘	5	import com.alibaba.fastjson.JSONObject;
	6	import com.iail.model.IAILModel;
4f1717	7	import com.iailab.module.model.common.enums.CommonConstant;
373ab1	8	import com.iailab.module.model.common.enums.OutResultType;
1178da	9	import com.iailab.module.model.common.exception.ModelResultErrorException;
69bd5e	10	import com.iailab.module.model.mcs.pre.entity.MmItemOutputEntity;
7fd198	11	import com.iailab.module.model.mcs.pre.entity.MmModelArithSettingsEntity;
潘	12	import com.iailab.module.model.mcs.pre.entity.MmPredictModelEntity;
7f0bcd	13	import com.iailab.module.model.mcs.pre.enums.ItemRunStatusEnum;
69bd5e	14	import com.iailab.module.model.mcs.pre.service.MmItemOutputService;
7fd198	15	import com.iailab.module.model.mcs.pre.service.MmModelArithSettingsService;
潘	16	import com.iailab.module.model.mdk.common.enums.TypeA;
	17	import com.iailab.module.model.mdk.common.exceptions.ModelInvokeException;
	18	import com.iailab.module.model.mdk.predict.PredictModelHandler;
	19	import com.iailab.module.model.mdk.sample.SampleConstructor;
	20	import com.iailab.module.model.mdk.sample.dto.SampleData;
	21	import com.iailab.module.model.mdk.vo.PredictResultVO;
45520a	22	import com.iailab.module.model.mpk.common.MdkConstant;
1a2b62	23	import com.iailab.module.model.mpk.common.utils.DllUtils;
7fd198	24	import lombok.extern.slf4j.Slf4j;
潘	25	import org.springframework.beans.factory.annotation.Autowired;
	26	import org.springframework.stereotype.Component;
	27
45520a	28	import java.util.Date;
D	29	import java.util.HashMap;
	30	import java.util.List;
	31	import java.util.Map;
7fd198	32
潘	33	/**
	34	* @author PanZhibao
	35	* @Description
	36	* @createTime 2024年09月01日
	37	*/
	38	@Slf4j
	39	@Component
	40	public class PredictModelHandlerImpl implements PredictModelHandler {
	41
	42	@Autowired
	43	private MmModelArithSettingsService mmModelArithSettingsService;
	44
	45	@Autowired
69bd5e	46	private MmItemOutputService mmItemOutputService;
7fd198	47
潘	48	@Autowired
	49	private SampleConstructor sampleConstructor;
	50
4f1717	51	/**
潘	52	* 根据模型预测,返回预测结果
	53	*
	54	* @param predictTime
	55	* @param predictModel
	56	* @return
	57	* @throws ModelInvokeException
	58	*/
7fd198	59	@Override
28c2db	60	public synchronized PredictResultVO predictByModel(Date predictTime, MmPredictModelEntity predictModel,String itemName,String itemNo) throws ModelInvokeException {
7fd198	61	PredictResultVO result = new PredictResultVO();
潘	62	if (predictModel == null) {
	63	throw new ModelInvokeException("modelEntity is null");
	64	}
	65	String modelId = predictModel.getId();
	66	try {
50084d	67	List<SampleData> sampleDataList = sampleConstructor.constructSample(TypeA.Predict.name(), modelId, predictTime, itemName);
7fd198	68	String modelPath = predictModel.getModelpath();
潘	69	if (modelPath == null) {
373ab1	70	log.info("模型路径不存在，modelId=" + modelId);
7fd198	71	return null;
潘	72	}
	73	IAILModel newModelBean = composeNewModelBean(predictModel);
	74	HashMap<String, Object> settings = getPredictSettingsByModelId(modelId);
45520a	75	// 校验setting必须有pyFile，否则可能导致程序崩溃
D	76	if (!settings.containsKey(MdkConstant.PY_FILE_KEY)) {
	77	throw new RuntimeException("模型设置参数缺少必要信息【" + MdkConstant.PY_FILE_KEY + "】,请重新上传模型！");
	78	}
	79
7fd198	80	if (settings == null) {
潘	81	log.error("模型setting不存在，modelId=" + modelId);
	82	return null;
	83	}
	84	int portLength = sampleDataList.size();
	85	Object[] param2Values = new Object[portLength + 2];
	86	for (int i = 0; i < portLength; i++) {
4f1717	87	param2Values[i] = sampleDataList.get(i).getMatrix();
7fd198	88	}
潘	89	param2Values[portLength] = newModelBean.getDataMap().get("models");
4f1717	90	param2Values[portLength + 1] = settings;
7fd198	91
595ccf	92	log.info("####################### 预测模型 "+ "【itemId:" + predictModel.getItemid() + ",itemName：" + itemName + ",itemNo：" + itemNo + "】 ##########################");
b82ba2	93	// JSONObject jsonObjNewModelBean = new JSONObject();
D	94	// jsonObjNewModelBean.put("newModelBean", newModelBean);
	95	// log.info(String.valueOf(jsonObjNewModelBean));
	96	// JSONObject jsonObjParam2Values = new JSONObject();
	97	// jsonObjParam2Values.put("param2Values", param2Values);
6957a3	98	log.info("参数： " + JSON.toJSONString(param2Values));
7fd198	99
潘	100	//IAILMDK.run
1a2b62	101	HashMap<String, Object> modelResult = DllUtils.run(newModelBean, param2Values, predictModel.getMpkprojectid());
73a05d	102	//打印结果
D	103	log.info("预测模型计算完成：modelId=" + modelId + ",modelName=" + predictModel.getMethodname() + "，modelResult=" + JSON.toJSONString(modelResult));
	104	//判断模型结果
4f1717	105	if (!modelResult.containsKey(CommonConstant.MDK_STATUS_CODE) \|\| !modelResult.containsKey(CommonConstant.MDK_RESULT) \|\|
潘	106	!modelResult.get(CommonConstant.MDK_STATUS_CODE).toString().equals(CommonConstant.MDK_STATUS_100)) {
1178da	107	throw new ModelResultErrorException("模型结果异常：" + modelResult);
b2aca2	108	}
4f1717	109	modelResult = (HashMap<String, Object>) modelResult.get(CommonConstant.MDK_RESULT);
7fd198	110
373ab1	111	List<MmItemOutputEntity> itemOutputList = mmItemOutputService.getByItemid(predictModel.getItemid());
a6e46f	112	Map<MmItemOutputEntity, double[]> predictMatrixs = new HashMap<>();
D	113	Map<MmItemOutputEntity, Double> predictDoubleValues = new HashMap<>();
373ab1	114	for (MmItemOutputEntity output : itemOutputList) {
潘	115	if (!modelResult.containsKey(output.getResultstr())) {
	116	continue;
	117	}
	118	OutResultType outResultType = OutResultType.getEumByCode(output.getResultType());
	119	switch (outResultType) {
	120	case D1:
	121	double[] temp1 = (double[]) modelResult.get(output.getResultstr());
	122	predictMatrixs.put(output, temp1);
	123	break;
	124	case D2:
	125	double[][] temp2 = (double[][]) modelResult.get(output.getResultstr());
	126	double[] tempColumn = new double[temp2.length];
	127	for (int i = 0; i < tempColumn.length; i++) {
	128	tempColumn[i] = temp2[i][output.getResultIndex()];
69bd5e	129	}
373ab1	130	predictMatrixs.put(output, tempColumn);
潘	131	break;
a6e46f	132	case D:
D	133	Double temp3 = (Double) modelResult.get(output.getResultstr());
	134	predictDoubleValues.put(output, temp3);
	135	break;
373ab1	136	default:
潘	137	break;
1a2b62	138	}
D	139	}
69bd5e	140	result.setPredictMatrixs(predictMatrixs);
a6e46f	141	result.setPredictDoubleValues(predictDoubleValues);
1a2b62	142	result.setModelResult(modelResult);
7fd198	143	result.setPredictTime(predictTime);
50084d	144	} catch (ModelResultErrorException ex) {
efdc38	145	// ex.printStackTrace();
D	146	log.error("模型结果异常", ex);
50084d	147	throw ex;
D	148	} catch (Exception ex) {
c9dd12	149	// log.error("调用发生异常，异常信息为:{0}", ex.getMessage());
efdc38	150	// ex.printStackTrace();
D	151	log.error("模型运行异常", ex);
ead005	152	throw new ModelInvokeException(ex.getMessage());
7fd198	153	}
潘	154	return result;
	155	}
	156
	157	/**
	158	* 构造IAILMDK.run()方法的newModelBean参数
	159	*
	160	* @param predictModel
	161	* @return
	162	*/
	163	private IAILModel composeNewModelBean(MmPredictModelEntity predictModel) {
	164	IAILModel newModelBean = new IAILModel();
	165	newModelBean.setClassName(predictModel.getClassname().trim());
	166	newModelBean.setMethodName(predictModel.getMethodname().trim());
	167	//构造参数类型
	168	String[] paArStr = predictModel.getModelparamstructure().trim().split(",");
	169	Class<?>[] paramsArray = new Class[paArStr.length];
	170	for (int i = 0; i < paArStr.length; i++) {
	171	if ("[[D".equals(paArStr[i])) {
	172	paramsArray[i] = double[][].class;
	173	} else if ("Map".equals(paArStr[i]) \|\| "java.util.HashMap".equals(paArStr[i])) {
	174	paramsArray[i] = HashMap.class;
	175	}
	176	}
	177	newModelBean.setParamsArray(paramsArray);
	178	HashMap<String, Object> dataMap = new HashMap<>();
	179	HashMap<String, String> models = new HashMap<>(1);
b2aca2	180	models.put("model_path", predictModel.getModelpath());
7fd198	181	dataMap.put("models", models);
潘	182	newModelBean.setDataMap(dataMap);
	183	return newModelBean;
	184	}
	185
	186	/**
	187	* 根据模型id获取参数map
	188	*
	189	* @param modelId
	190	* @return
	191	*/
	192	private HashMap<String, Object> getPredictSettingsByModelId(String modelId) {
	193	List<MmModelArithSettingsEntity> list = mmModelArithSettingsService.getByModelId(modelId);
	194	HashMap<String, Object> result = new HashMap<>();
	195	for (MmModelArithSettingsEntity entry : list) {
	196	String valueType = entry.getValuetype().trim(); //去除两端空格
	197	if ("int".equals(valueType)) {
	198	int value = Integer.parseInt(entry.getValue());
	199	result.put(entry.getKey(), value);
	200	} else if ("double".equals(valueType)) {
	201	double value = Double.parseDouble(entry.getValue());
	202	result.put(entry.getKey(), value);
	203	} else if ("string".equals(valueType)) {
	204	String value = entry.getValue();
	205	result.put(entry.getKey(), value);
	206	} else if ("decimalArray".equals(valueType)) {
	207	JSONArray valueArray = JSONArray.parseArray(entry.getValue());
	208	double[] value = new double[valueArray.size()];
	209	for (int i = 0; i < valueArray.size(); i++) {
	210	value[i] = Double.parseDouble(valueArray.get(i).toString());
	211	}
	212	result.put(entry.getKey(), value);
	213	} else if ("decimal".equals(valueType)) {
	214	double value = Double.parseDouble(entry.getValue());
	215	result.put(entry.getKey(), value);
	216	}
	217	}
	218	return result;
	219	}
	220	}