Question

我已经在Java代码中使用autoML构建了StackedEnsembleModel。但是，当我尝试将模型导出到mojo zip文件时，出现了“ water.exceptions.H2ONotFoundArgumentException：无法找到版本3的架构并键入：StackedEnsembleModel”的异常。

initH2OCloud();
try {
    registerH2OAlgorithm();
    AutoMLBuildSpec autoMLBuildSpec = new AutoMLBuildSpec();
    Frame trainFrame = importAndParseFeatureFileToH2O("/home/data/MODEL.csv");

    autoMLBuildSpec.input_spec.training_frame = trainFrame._key;
    autoMLBuildSpec.input_spec.response_column = "level";

    autoMLBuildSpec.build_models.exclude_algos = new Algo[2];
    autoMLBuildSpec.build_models.exclude_algos[0] = Algo.DeepLearning;
    autoMLBuildSpec.build_models.exclude_algos[1] = Algo.XGBoost;
    autoMLBuildSpec.build_control.keep_cross_validation_models = true;
    autoMLBuildSpec.build_control.keep_cross_validation_predictions = true;
    autoMLBuildSpec.build_control.keep_cross_validation_fold_assignment = true;
    autoMLBuildSpec.build_control.nfolds = 5;
    autoMLBuildSpec.build_control.project_name = "test.prj";
    autoMLBuildSpec.build_control.stopping_criteria.set_max_runtime_secs(300);
    AutoML aml = AutoML.makeAutoML(Key.make(), new Date(), autoMLBuildSpec);
    logger.info("Begin AutoML...");
    long beginTick = Calendar.getInstance().getTime().getTime();
    AutoML.startAutoML(autoMLBuildSpec).get();
    long usedTick = Calendar.getInstance().getTime().getTime() - beginTick;
    //save the leader model
    String modelSavePath = "/home/data/tmp";
    logger.info("best model: {}, RMSE: {}, MAE: {}, RSLE: {}。",
    aml.leader()._key.toString(),
    bigDecimalToString(new BigDecimal(Math.sqrt(aml.leader().mse())), 12),
    bigDecimalToString(new BigDecimal(Math.sqrt(aml.leader().mae())), 12),
    bigDecimalToString(new BigDecimal(Math.sqrt(aml.leader().rmsle())), 12));
    logger.info("used {} seconds", usedTick/1000L);
    logger.info("export to mojo...");
    try {
        aml.leader().getMojo().writeTo(new FileOutputStream(new File(modelSavePath + File.separator + aml.leader()._key + ".zip")));
        logger.info("export finished.");
    }
    catch (Throwable t) {
        logger.error("got exception", t);
    }
    logger.info("aml.leaderboard: ");
    logger.info(aml.leaderboard());
}
finally {
    stopH2OCloud();
}

当我调用aml.leader（）。getMojo（）。writeTo（）方法时引发的异常。我正在使用h2o 3.22.0.1。预先感谢。

Answer 1

尝试使用aml.leader().getMojo().writeTo()代替aml.leader().exportMojo(...)。

尝试导出Mojo模型时出现“ water.exceptions.H2ONotFoundArgumentException”

1 个答案: