[jira] [Created] (SYSTEMML-1849) Can't obtain scores from l2-svm-predict.dml using JMLC

[Deron Eriksson (JIRA)]

Deron Eriksson created SYSTEMML-1849:
----------------------------------------

             Summary: Can't obtain scores from l2-svm-predict.dml using JMLC
                 Key: SYSTEMML-1849
                 URL: https://issues.apache.org/jira/browse/SYSTEMML-1849
             Project: SystemML
          Issue Type: Bug
          Components: Compiler, Runtime
            Reporter: Deron Eriksson


Attempting to obtain the {{scores}} variable in {{l2-svm-predict.dml}} using 
JMLC gives:
{code}
Exception in thread "main" org.apache.sysml.api.DMLException: Non-existent 
output variable: scores
        at 
org.apache.sysml.api.jmlc.ResultVariables.getMatrixBlock(ResultVariables.java:88)
        at 
org.apache.sysml.api.jmlc.ResultVariables.getMatrix(ResultVariables.java:74)
{code}

This appears to be introduced by 
https://github.com/apache/systemml/commit/4b81d0dda6583f0ae96eaa7aa5832005ae5fa8a9

Example code:
{code}
        public static void jmlcL2SVM() throws Exception {

                Connection conn = new Connection();

                String dml = conn.readScript("scripts/algorithms/l2-svm.dml");
                PreparedScript l2svm = conn.prepareScript(dml, new String[] { 
"X", "Y", "fmt", "Log" },
                                new String[] { "w", "debug_str" }, false);

                double[][] trainData = new double[150][3];
                for (int i = 0; i < 150; i++) {
                        int one = ThreadLocalRandom.current().nextInt(0, 101);
                        int two = ThreadLocalRandom.current().nextInt(0, 101);
                        int three = ThreadLocalRandom.current().nextInt(0, 101);
                        double[] row = new double[] { one, two, three };
                        trainData[i] = row;
                }

                l2svm.setMatrix("X", trainData);
                log.debug(displayMatrix(trainData));

                double[][] trainLabels = new double[150][1];
                for (int i = 0; i < 150; i++) {
                        int one = ThreadLocalRandom.current().nextInt(1, 3);
                        double[] row = new double[] { one };
                        trainLabels[i] = row;
                }
                l2svm.setMatrix("Y", trainLabels);
                log.debug(displayMatrix(trainLabels));

                l2svm.setScalar("fmt", "csv");

                l2svm.setScalar("Log", "temp/l2-svm-log.csv");

                ResultVariables l2svmResults = l2svm.executeScript();
                double[][] model = l2svmResults.getMatrix("w");
                log.debug("MODEL:");
                log.debug(displayMatrix(model));
                String debugString = l2svmResults.getString("debug_str");
                log.debug("DEBUG STRING:");
                log.debug(debugString);

                String s = 
conn.readScript("scripts/algorithms/l2-svm-predict.dml");
                Map<String, String> m = new HashMap<String, String>();
                m.put("$Y", "\"temp/haberman.test.labels.csv\"");
                m.put("$confusion", "\"temp/l2-svm-confusion.csv\"");

                PreparedScript l2svmPredict = conn.prepareScript(s, m, new 
String[] { "X", "y", "w", "fmt" },
                                new String[] { "scores", "confusion_mat" }, 
false);

                double[][] testData = new double[150][3];
                for (int i = 0; i < 150; i++) {
                        int one = ThreadLocalRandom.current().nextInt(0, 101);
                        int two = ThreadLocalRandom.current().nextInt(0, 101);
                        int three = ThreadLocalRandom.current().nextInt(0, 101);
                        double[] row = new double[] { one, two, three };
                        testData[i] = row;
                }
                l2svmPredict.setMatrix("X", testData);

                double[][] testLabels = new double[150][1];
                for (int i = 0; i < 150; i++) {
                        int one = ThreadLocalRandom.current().nextInt(1, 3);
                        double[] row = new double[] { one };
                        testLabels[i] = row;
                }
                l2svmPredict.setMatrix("y", testLabels);

                l2svmPredict.setMatrix("w", model);

                l2svmPredict.setScalar("fmt", "csv");

                ResultVariables l2svmPredictResults = 
l2svmPredict.executeScript();
                double[][] scores = l2svmPredictResults.getMatrix("scores");
                log.debug("SCORES:");
                log.debug(displayMatrix(scores));

                double[][] confusionMatrix = 
l2svmPredictResults.getMatrix("confusion_mat");
                log.debug("CONFUSION MATRIX:");
                log.debug(displayMatrix(confusionMatrix));

                conn.close();
        }
{code}





--
This message was sent by Atlassian JIRA
(v6.4.14#64029)