iailab-plat.git

潘志宝

2025-02-25 b3674cefbeb8ffaf49f96d0c8fe7a6694ed86eec

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.iail.model.IAILModel;
b3674c	6	import com.iailab.module.model.mdk.vo.StAdjustDeviationDTO;
401492	7	import com.iailab.module.model.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;
69bd5e	13	import com.iailab.module.model.mcs.pre.service.MmItemOutputService;
7fd198	14	import com.iailab.module.model.mcs.pre.service.MmModelArithSettingsService;
潘	15	import com.iailab.module.model.mdk.common.enums.TypeA;
	16	import com.iailab.module.model.mdk.common.exceptions.ModelInvokeException;
	17	import com.iailab.module.model.mdk.predict.PredictModelHandler;
	18	import com.iailab.module.model.mdk.sample.SampleConstructor;
	19	import com.iailab.module.model.mdk.sample.dto.SampleData;
	20	import com.iailab.module.model.mdk.vo.PredictResultVO;
45520a	21	import com.iailab.module.model.mpk.common.MdkConstant;
1a2b62	22	import com.iailab.module.model.mpk.common.utils.DllUtils;
7fd198	23	import lombok.extern.slf4j.Slf4j;
潘	24	import org.springframework.beans.factory.annotation.Autowired;
	25	import org.springframework.stereotype.Component;
	26
45520a	27	import java.util.Date;
D	28	import java.util.HashMap;
	29	import java.util.List;
	30	import java.util.Map;
7fd198	31
潘	32	/**
	33	* @author PanZhibao
	34	* @Description
	35	* @createTime 2024年09月01日
	36	*/
	37	@Slf4j
	38	@Component
	39	public class PredictModelHandlerImpl implements PredictModelHandler {
	40
	41	@Autowired
	42	private MmModelArithSettingsService mmModelArithSettingsService;
	43
	44	@Autowired
69bd5e	45	private MmItemOutputService mmItemOutputService;
7fd198	46
潘	47	@Autowired
	48	private SampleConstructor sampleConstructor;
	49
4f1717	50	/**
潘	51	* 根据模型预测,返回预测结果
	52	*
	53	* @param predictTime
	54	* @param predictModel
	55	* @return
	56	* @throws ModelInvokeException
	57	*/
7fd198	58	@Override
b3674c	59	public PredictResultVO predictByModel(Date predictTime, MmPredictModelEntity predictModel, String itemName, String itemNo) throws ModelInvokeException {
7fd198	60	PredictResultVO result = new PredictResultVO();
潘	61	if (predictModel == null) {
	62	throw new ModelInvokeException("modelEntity is null");
	63	}
	64	String modelId = predictModel.getId();
	65	try {
81ce77	66	List<SampleData> sampleDataList = sampleConstructor.constructSample(TypeA.Predict.name(), modelId, predictTime, itemName, new HashMap<>());
7fd198	67	String modelPath = predictModel.getModelpath();
潘	68	if (modelPath == null) {
373ab1	69	log.info("模型路径不存在，modelId=" + modelId);
7fd198	70	return null;
潘	71	}
	72	IAILModel newModelBean = composeNewModelBean(predictModel);
	73	HashMap<String, Object> settings = getPredictSettingsByModelId(modelId);
45520a	74	// 校验setting必须有pyFile，否则可能导致程序崩溃
D	75	if (!settings.containsKey(MdkConstant.PY_FILE_KEY)) {
b3674c	76	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！");
45520a	77	}
D	78
7fd198	79	if (settings == null) {
潘	80	log.error("模型setting不存在，modelId=" + modelId);
	81	return null;
	82	}
	83	int portLength = sampleDataList.size();
	84	Object[] param2Values = new Object[portLength + 2];
	85	for (int i = 0; i < portLength; i++) {
4f1717	86	param2Values[i] = sampleDataList.get(i).getMatrix();
7fd198	87	}
潘	88	param2Values[portLength] = newModelBean.getDataMap().get("models");
4f1717	89	param2Values[portLength + 1] = settings;
7fd198	90
b3674c	91	log.info("####################### 预测模型 " + "【itemId:" + predictModel.getItemid() + ",itemName：" + itemName + ",itemNo：" + itemNo + "】 ##########################");
6957a3	92	log.info("参数： " + JSON.toJSONString(param2Values));
7fd198	93
潘	94	//IAILMDK.run
1a2b62	95	HashMap<String, Object> modelResult = DllUtils.run(newModelBean, param2Values, predictModel.getMpkprojectid());
73a05d	96	//打印结果
D	97	log.info("预测模型计算完成：modelId=" + modelId + ",modelName=" + predictModel.getMethodname() + "，modelResult=" + JSON.toJSONString(modelResult));
	98	//判断模型结果
4f1717	99	if (!modelResult.containsKey(CommonConstant.MDK_STATUS_CODE) \|\| !modelResult.containsKey(CommonConstant.MDK_RESULT) \|\|
潘	100	!modelResult.get(CommonConstant.MDK_STATUS_CODE).toString().equals(CommonConstant.MDK_STATUS_100)) {
1178da	101	throw new ModelResultErrorException("模型结果异常：" + modelResult);
b2aca2	102	}
4f1717	103	modelResult = (HashMap<String, Object>) modelResult.get(CommonConstant.MDK_RESULT);
7fd198	104
373ab1	105	List<MmItemOutputEntity> itemOutputList = mmItemOutputService.getByItemid(predictModel.getItemid());
a6e46f	106	Map<MmItemOutputEntity, double[]> predictMatrixs = new HashMap<>();
373ab1	107	for (MmItemOutputEntity output : itemOutputList) {
潘	108	if (!modelResult.containsKey(output.getResultstr())) {
	109	continue;
	110	}
	111	OutResultType outResultType = OutResultType.getEumByCode(output.getResultType());
	112	switch (outResultType) {
	113	case D1:
	114	double[] temp1 = (double[]) modelResult.get(output.getResultstr());
	115	predictMatrixs.put(output, temp1);
	116	break;
	117	case D2:
	118	double[][] temp2 = (double[][]) modelResult.get(output.getResultstr());
	119	double[] tempColumn = new double[temp2.length];
	120	for (int i = 0; i < tempColumn.length; i++) {
	121	tempColumn[i] = temp2[i][output.getResultIndex()];
69bd5e	122	}
373ab1	123	predictMatrixs.put(output, tempColumn);
潘	124	break;
a6e46f	125	case D:
D	126	Double temp3 = (Double) modelResult.get(output.getResultstr());
51e472	127	predictMatrixs.put(output, new double[]{temp3});
a6e46f	128	break;
373ab1	129	default:
潘	130	break;
1a2b62	131	}
D	132	}
69bd5e	133	result.setPredictMatrixs(predictMatrixs);
1a2b62	134	result.setModelResult(modelResult);
7fd198	135	result.setPredictTime(predictTime);
50084d	136	} catch (ModelResultErrorException ex) {
efdc38	137	// ex.printStackTrace();
D	138	log.error("模型结果异常", ex);
50084d	139	throw ex;
D	140	} catch (Exception ex) {
c9dd12	141	// log.error("调用发生异常，异常信息为:{0}", ex.getMessage());
efdc38	142	// ex.printStackTrace();
D	143	log.error("模型运行异常", ex);
ead005	144	throw new ModelInvokeException(ex.getMessage());
7fd198	145	}
潘	146	return result;
	147	}
	148
	149	/**
fdcde1	150	* 预测，模拟调整
潘	151	*
	152	* @param predictTime
	153	* @param predictModel
	154	* @param itemName
	155	* @param itemNo
b3674c	156	* @param deviationList
fdcde1	157	* @return
潘	158	* @throws ModelInvokeException
	159	*/
	160	@Override
b3674c	161	public PredictResultVO predictByModel(Date predictTime, MmPredictModelEntity predictModel, String itemName, String itemNo, List<StAdjustDeviationDTO> deviationList) throws ModelInvokeException {
fdcde1	162	PredictResultVO result = new PredictResultVO();
潘	163	if (predictModel == null) {
	164	throw new ModelInvokeException("modelEntity is null");
	165	}
	166	String modelId = predictModel.getId();
	167	try {
b3674c	168	List<SampleData> sampleDataList = sampleConstructor.constructSample(TypeA.Predict.name(), modelId, predictTime, itemName, new HashMap<>(), deviationList);
fdcde1	169	String modelPath = predictModel.getModelpath();
潘	170	if (modelPath == null) {
	171	log.info("模型路径不存在，modelId=" + modelId);
	172	return null;
	173	}
	174	IAILModel newModelBean = composeNewModelBean(predictModel);
	175	HashMap<String, Object> settings = getPredictSettingsByModelId(modelId);
	176	// 校验setting必须有pyFile，否则可能导致程序崩溃
	177	if (!settings.containsKey(MdkConstant.PY_FILE_KEY)) {
b3674c	178	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！");
fdcde1	179	}
潘	180
	181	if (settings == null) {
	182	log.error("模型setting不存在，modelId=" + modelId);
	183	return null;
	184	}
	185	int portLength = sampleDataList.size();
	186	Object[] param2Values = new Object[portLength + 2];
	187	for (int i = 0; i < portLength; i++) {
	188	param2Values[i] = sampleDataList.get(i).getMatrix();
	189	}
	190	param2Values[portLength] = newModelBean.getDataMap().get("models");
	191	param2Values[portLength + 1] = settings;
	192
b3674c	193	log.info("####################### 模拟调整 " + "【itemId:" + predictModel.getItemid() + ",itemName：" + itemName + ",itemNo：" + itemNo + "】 ##########################");
fdcde1	194	log.info("参数： " + JSON.toJSONString(param2Values));
潘	195
	196	//IAILMDK.run
	197	HashMap<String, Object> modelResult = DllUtils.run(newModelBean, param2Values, predictModel.getMpkprojectid());
	198	//打印结果
	199	log.info("预测模型计算完成：modelId=" + modelId + ",modelName=" + predictModel.getMethodname() + "，modelResult=" + JSON.toJSONString(modelResult));
	200	//判断模型结果
	201	if (!modelResult.containsKey(CommonConstant.MDK_STATUS_CODE) \|\| !modelResult.containsKey(CommonConstant.MDK_RESULT) \|\|
	202	!modelResult.get(CommonConstant.MDK_STATUS_CODE).toString().equals(CommonConstant.MDK_STATUS_100)) {
	203	throw new ModelResultErrorException("模型结果异常：" + modelResult);
	204	}
	205	modelResult = (HashMap<String, Object>) modelResult.get(CommonConstant.MDK_RESULT);
	206
	207	List<MmItemOutputEntity> itemOutputList = mmItemOutputService.getByItemid(predictModel.getItemid());
	208	Map<MmItemOutputEntity, double[]> predictMatrixs = new HashMap<>();
	209	for (MmItemOutputEntity output : itemOutputList) {
	210	if (!modelResult.containsKey(output.getResultstr())) {
	211	continue;
	212	}
	213	OutResultType outResultType = OutResultType.getEumByCode(output.getResultType());
	214	switch (outResultType) {
	215	case D1:
	216	double[] temp1 = (double[]) modelResult.get(output.getResultstr());
	217	predictMatrixs.put(output, temp1);
	218	break;
	219	case D2:
	220	double[][] temp2 = (double[][]) modelResult.get(output.getResultstr());
	221	double[] tempColumn = new double[temp2.length];
	222	for (int i = 0; i < tempColumn.length; i++) {
	223	tempColumn[i] = temp2[i][output.getResultIndex()];
	224	}
	225	predictMatrixs.put(output, tempColumn);
	226	break;
	227	case D:
	228	Double temp3 = (Double) modelResult.get(output.getResultstr());
	229	predictMatrixs.put(output, new double[]{temp3});
	230	break;
	231	default:
	232	break;
	233	}
	234	}
	235	result.setPredictMatrixs(predictMatrixs);
	236	result.setModelResult(modelResult);
	237	result.setPredictTime(predictTime);
	238	} catch (ModelResultErrorException ex) {
	239	log.error("模型结果异常", ex);
	240	throw ex;
	241	} catch (Exception ex) {
	242	log.error("调用发生异常，异常信息为:{0}", ex.getMessage());
	243	throw new ModelInvokeException(ex.getMessage());
	244	}
	245	return result;
	246	}
b3674c	247
fdcde1	248	/**
7fd198	249	* 构造IAILMDK.run()方法的newModelBean参数
潘	250	*
	251	* @param predictModel
	252	* @return
	253	*/
	254	private IAILModel composeNewModelBean(MmPredictModelEntity predictModel) {
	255	IAILModel newModelBean = new IAILModel();
	256	newModelBean.setClassName(predictModel.getClassname().trim());
	257	newModelBean.setMethodName(predictModel.getMethodname().trim());
	258	//构造参数类型
	259	String[] paArStr = predictModel.getModelparamstructure().trim().split(",");
	260	Class<?>[] paramsArray = new Class[paArStr.length];
	261	for (int i = 0; i < paArStr.length; i++) {
	262	if ("[[D".equals(paArStr[i])) {
	263	paramsArray[i] = double[][].class;
	264	} else if ("Map".equals(paArStr[i]) \|\| "java.util.HashMap".equals(paArStr[i])) {
	265	paramsArray[i] = HashMap.class;
	266	}
	267	}
	268	newModelBean.setParamsArray(paramsArray);
	269	HashMap<String, Object> dataMap = new HashMap<>();
	270	HashMap<String, String> models = new HashMap<>(1);
b2aca2	271	models.put("model_path", predictModel.getModelpath());
7fd198	272	dataMap.put("models", models);
潘	273	newModelBean.setDataMap(dataMap);
	274	return newModelBean;
	275	}
	276
	277	/**
	278	* 根据模型id获取参数map
	279	*
	280	* @param modelId
	281	* @return
	282	*/
	283	private HashMap<String, Object> getPredictSettingsByModelId(String modelId) {
	284	List<MmModelArithSettingsEntity> list = mmModelArithSettingsService.getByModelId(modelId);
	285	HashMap<String, Object> result = new HashMap<>();
	286	for (MmModelArithSettingsEntity entry : list) {
	287	String valueType = entry.getValuetype().trim(); //去除两端空格
	288	if ("int".equals(valueType)) {
	289	int value = Integer.parseInt(entry.getValue());
	290	result.put(entry.getKey(), value);
	291	} else if ("double".equals(valueType)) {
	292	double value = Double.parseDouble(entry.getValue());
	293	result.put(entry.getKey(), value);
	294	} else if ("string".equals(valueType)) {
	295	String value = entry.getValue();
	296	result.put(entry.getKey(), value);
	297	} else if ("decimalArray".equals(valueType)) {
	298	JSONArray valueArray = JSONArray.parseArray(entry.getValue());
	299	double[] value = new double[valueArray.size()];
	300	for (int i = 0; i < valueArray.size(); i++) {
	301	value[i] = Double.parseDouble(valueArray.get(i).toString());
	302	}
	303	result.put(entry.getKey(), value);
	304	} else if ("decimal".equals(valueType)) {
	305	double value = Double.parseDouble(entry.getValue());
	306	result.put(entry.getKey(), value);
	307	}
	308	}
	309	return result;
	310	}
	311	}