iailab-module-model/iailab-module-model-biz/src/main/java/com/iailab/module/model/mdk/predict/impl/PredictModelHandlerImpl.java

@@ -2,28 +2,30 @@

import com.alibaba.fastjson.JSONArray;
import com.alibaba.fastjson.JSONObject;
import com.iail.IAILMDK;
import com.iail.model.IAILModel;
import com.iailab.module.model.common.enums.CommonConstant;
import com.iailab.module.model.common.enums.OutResultType;
import com.iailab.module.model.mcs.pre.entity.MmItemOutputEntity;
import com.iailab.module.model.mcs.pre.entity.MmModelArithSettingsEntity;
import com.iailab.module.model.mcs.pre.entity.MmModelResultstrEntity;
import com.iailab.module.model.mcs.pre.entity.MmPredictModelEntity;
import com.iailab.module.model.mcs.pre.service.MmItemOutputService;
import com.iailab.module.model.mcs.pre.service.MmModelArithSettingsService;
import com.iailab.module.model.mcs.pre.service.MmModelResultstrService;
import com.iailab.module.model.mdk.common.enums.TypeA;
import com.iailab.module.model.mdk.common.exceptions.ModelInvokeException;
import com.iailab.module.model.mdk.predict.PredictModelHandler;
import com.iailab.module.model.mdk.sample.SampleConstructor;
import com.iailab.module.model.mdk.sample.dto.SampleData;
import com.iailab.module.model.mdk.vo.PredictResultVO;
import com.iailab.module.model.mpk.common.MdkConstant;
import com.iailab.module.model.mpk.common.utils.DllUtils;
import lombok.extern.slf4j.Slf4j;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.stereotype.Component;

import java.util.ArrayList;
import java.util.Date;
import java.util.HashMap;
import java.util.List;
import java.util.Map;

/**
 * @author PanZhibao
@@ -38,28 +40,40 @@
    private MmModelArithSettingsService mmModelArithSettingsService;

    @Autowired
    private MmModelResultstrService mmModelResultstrService;
    private MmItemOutputService mmItemOutputService;

    @Autowired
    private SampleConstructor sampleConstructor;

    /**
     * 根据模型预测,返回预测结果
     *
     * @param predictTime
     * @param predictModel
     * @return
     * @throws ModelInvokeException
     */
    @Override
    public PredictResultVO predictByModel(Date predictTime, MmPredictModelEntity predictModel) throws ModelInvokeException {
    public synchronized PredictResultVO predictByModel(Date predictTime, MmPredictModelEntity predictModel) throws ModelInvokeException {
        PredictResultVO result = new PredictResultVO();
        if (predictModel == null) {
            throw new ModelInvokeException("modelEntity is null");
        }
        String modelId = predictModel.getId();

        try {
            List<SampleData> sampleDataList = sampleConstructor.constructSample(TypeA.Predict.name(), modelId, predictTime);
            String modelPath = predictModel.getModelpath();
            if (modelPath == null) {
                System.out.println("模型路径不存在，modelId=" + modelId);
                log.info("模型路径不存在，modelId=" + modelId);
                return null;
            }
            IAILModel newModelBean = composeNewModelBean(predictModel);
            HashMap<String, Object> settings = getPredictSettingsByModelId(modelId);
            // 校验setting必须有pyFile，否则可能导致程序崩溃
            if (!settings.containsKey(MdkConstant.PY_FILE_KEY)) {
                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！");
            }

            if (settings == null) {
                log.error("模型setting不存在，modelId=" + modelId);
                return null;
@@ -67,52 +81,69 @@
            int portLength = sampleDataList.size();
            Object[] param2Values = new Object[portLength + 2];
            for (int i = 0; i < portLength; i++) {
                param2Values[i]=sampleDataList.get(i).getMatrix();
                param2Values[i] = sampleDataList.get(i).getMatrix();
            }
            param2Values[portLength] = newModelBean.getDataMap().get("models");
            param2Values[portLength+1] = settings;
            param2Values[portLength + 1] = settings;

            log.info("#######################预测模型 " + predictModel.getItemid() + " ##########################");
            JSONObject jsonObjNewModelBean = new JSONObject();
            jsonObjNewModelBean.put("newModelBean", newModelBean);
            log.info(String.valueOf(jsonObjNewModelBean));
            JSONObject jsonObjParam2Values = new JSONObject();
            jsonObjParam2Values.put("param2Values", param2Values);
            log.info(String.valueOf(jsonObjParam2Values));
            log.info("####################### 预测模型 "+ "【itemId:" + predictModel.getItemid() + ",modelName" + predictModel.getMethodname() + "】 ##########################");
//            JSONObject jsonObjNewModelBean = new JSONObject();
//            jsonObjNewModelBean.put("newModelBean", newModelBean);
//            log.info(String.valueOf(jsonObjNewModelBean));
//            JSONObject jsonObjParam2Values = new JSONObject();
//            jsonObjParam2Values.put("param2Values", param2Values);
//            log.info(String.valueOf(jsonObjParam2Values));

            //IAILMDK.run
//            HashMap<String, Object> modelResult = IAILMDK.run(newModelBean, param2Values);
            HashMap<String, Object> modelResult = DllUtils.run(newModelBean, param2Values, predictModel.getMpkprojectid());
            if(!modelResult.containsKey("status_code") || !modelResult.containsKey("result") || Integer.parseInt(modelResult.get("status_code").toString()) != 100) {
            if (!modelResult.containsKey(CommonConstant.MDK_STATUS_CODE) || !modelResult.containsKey(CommonConstant.MDK_RESULT) ||
                    !modelResult.get(CommonConstant.MDK_STATUS_CODE).toString().equals(CommonConstant.MDK_STATUS_100)) {
                throw new RuntimeException("模型结果异常：" + modelResult);
            }

            modelResult = (HashMap<String, Object>) modelResult.get("result");
            modelResult = (HashMap<String, Object>) modelResult.get(CommonConstant.MDK_RESULT);
            //打印结果
            log.info("预测模型计算完成：modelId=" + modelId + ",modelName" + predictModel.getMethodname());
            JSONObject jsonObjResult = new JSONObject();
            jsonObjResult.put("result", modelResult);
            log.info(String.valueOf(jsonObjResult));

            MmModelResultstrEntity modelResultstr = mmModelResultstrService.getInfo(predictModel.getResultstrid());
            log.info("模型计算完成：modelId=" + modelId + result);
            if (modelResult.containsKey(modelResultstr.getResultstr())) {
                Double[][] temp = (Double[][]) modelResult.get(modelResultstr.getResultstr());
                double[][] temp1 = new double[temp.length][temp[0].length];
                for (int i = 0; i < temp.length; i++) {
                    for (int j = 0; j < temp[i].length; j++) {
                        temp1[i][j] = temp[i][j].doubleValue();
                    }
            List<MmItemOutputEntity> itemOutputList = mmItemOutputService.getByItemid(predictModel.getItemid());
            Map<MmItemOutputEntity, double[]> predictMatrixs = new HashMap<>();
            Map<MmItemOutputEntity, Double> predictDoubleValues = new HashMap<>();
            for (MmItemOutputEntity output : itemOutputList) {
                if (!modelResult.containsKey(output.getResultstr())) {
                    continue;
                }
                result.setPredictMatrix(temp1);
                OutResultType outResultType = OutResultType.getEumByCode(output.getResultType());
                switch (outResultType) {
                    case D1:
                        double[] temp1 = (double[]) modelResult.get(output.getResultstr());
                        predictMatrixs.put(output, temp1);
                        break;
                    case D2:
                        double[][] temp2 = (double[][]) modelResult.get(output.getResultstr());
                        double[] tempColumn = new double[temp2.length];
                        for (int i = 0; i < tempColumn.length; i++) {
                            tempColumn[i] = temp2[i][output.getResultIndex()];
                        }
                        predictMatrixs.put(output, tempColumn);
                        break;
                    case D:
                        Double temp3 = (Double) modelResult.get(output.getResultstr());
                        predictDoubleValues.put(output, temp3);
                        break;
                    default:
                        break;
                }
            }
            result.setPredictMatrixs(predictMatrixs);
            result.setPredictDoubleValues(predictDoubleValues);
            result.setModelResult(modelResult);
            result.setPredictTime(predictTime);
        } catch (Exception ex) {
            log.error("IAILModel对象构造失败，modelId=" + modelId);
            log.error(ex.getMessage());
            log.error("调用发生异常，异常信息为:{}" , ex);
            log.error("调用发生异常，异常信息为:{}", ex);
            ex.printStackTrace();

            throw new ModelInvokeException(ex.getMessage());
        }
        return result;
    }