Examples with TwoDimTable water.util.TwoDimTable used on opensource projects

Example 16 with TwoDimTable

use of water.util.TwoDimTable in project h2o-3 by h2oai.

the class TwoDimTableTest method run0.

@Test
public void run0() {
    TwoDimTable table = new TwoDimTable("My foo bar table", null, new String[4], new String[] { "DoubleValue", "S2", "My Terrible Percent Value" }, new String[] { "double", "string", "double" }, new String[] { "%5.8e", "%s", "%5.8g %%" }, "", new String[][] { new String[] { null, "One", null }, new String[] { null, null, null }, new String[] { null, "Three", null }, new String[] { null, "FooBar", null } }, new double[][] { new double[] { 1.123, emptyDouble, 3200034.00001 }, new double[] { 123.34, emptyDouble, 1.0 }, new double[] { emptyDouble, emptyDouble, 3234.00001 }, new double[] { 3.33420923423423, emptyDouble, 3.40234234 } });
    String ts = table.toString();
    assertTrue(ts.length() > 0);
    Log.info(ts);
    String json = new TwoDimTableV3().fillFromImpl(table).toJsonString();
    Log.info(json);
}

Example 17 with TwoDimTable

use of water.util.TwoDimTable in project h2o-3 by h2oai.

the class ScoringInfo method createScoringHistoryTable.

/**
   * Create a TwoDimTable to display the scoring history from an array of scoringInfo.
   * @param scoringInfos array of ScoringInfo to render
   * @param hasValidation do we have validation metrics?
   * @param hasCrossValidation do we have cross-validation metrics?
   * @param modelCategory the category for the model or models
   * @param isAutoencoder is the model or are the models autoencoders?
   * @return
   */
public static TwoDimTable createScoringHistoryTable(ScoringInfo[] scoringInfos, boolean hasValidation, boolean hasCrossValidation, ModelCategory modelCategory, boolean isAutoencoder) {
    boolean hasEpochs = (scoringInfos instanceof HasEpochs[]);
    boolean hasSamples = (scoringInfos instanceof HasSamples[]);
    boolean hasIterations = (scoringInfos instanceof HasIterations[]);
    boolean isClassifier = (modelCategory == ModelCategory.Binomial || modelCategory == ModelCategory.Multinomial);
    List<String> colHeaders = new ArrayList<>();
    List<String> colTypes = new ArrayList<>();
    List<String> colFormat = new ArrayList<>();
    colHeaders.add("Timestamp");
    colTypes.add("string");
    colFormat.add("%s");
    colHeaders.add("Duration");
    colTypes.add("string");
    colFormat.add("%s");
    if (hasSamples) {
        colHeaders.add("Training Speed");
        colTypes.add("string");
        colFormat.add("%s");
    }
    if (hasEpochs) {
        colHeaders.add("Epochs");
        colTypes.add("double");
        colFormat.add("%.5f");
    }
    if (hasIterations) {
        colHeaders.add("Iterations");
        colTypes.add("int");
        colFormat.add("%d");
    }
    if (hasSamples) {
        colHeaders.add("Samples");
        colTypes.add("double");
        colFormat.add("%f");
    }
    colHeaders.add("Training RMSE");
    colTypes.add("double");
    colFormat.add("%.5f");
    if (modelCategory == ModelCategory.Regression) {
        colHeaders.add("Training Deviance");
        colTypes.add("double");
        colFormat.add("%.5f");
    }
    if (modelCategory == ModelCategory.Regression) {
        colHeaders.add("Training MAE");
        colTypes.add("double");
        colFormat.add("%.5f");
    }
    if (isClassifier) {
        colHeaders.add("Training LogLoss");
        colTypes.add("double");
        colFormat.add("%.5f");
    }
    if (modelCategory == ModelCategory.Binomial) {
        colHeaders.add("Training AUC");
        colTypes.add("double");
        colFormat.add("%.5f");
    }
    if (modelCategory == ModelCategory.Binomial) {
        colHeaders.add("Training Lift");
        colTypes.add("double");
        colFormat.add("%.5f");
    }
    if (isClassifier) {
        colHeaders.add("Training Classification Error");
        colTypes.add("double");
        colFormat.add("%.5f");
    }
    if (modelCategory == ModelCategory.AutoEncoder) {
        colHeaders.add("Training MSE");
        colTypes.add("double");
        colFormat.add("%.5f");
    }
    if (hasValidation) {
        colHeaders.add("Validation RMSE");
        colTypes.add("double");
        colFormat.add("%.5f");
        if (modelCategory == ModelCategory.Regression) {
            colHeaders.add("Validation Deviance");
            colTypes.add("double");
            colFormat.add("%.5f");
            colHeaders.add("Validation MAE");
            colTypes.add("double");
            colFormat.add("%.5f");
        }
        if (isClassifier) {
            colHeaders.add("Validation LogLoss");
            colTypes.add("double");
            colFormat.add("%.5f");
        }
        if (modelCategory == ModelCategory.Binomial) {
            colHeaders.add("Validation AUC");
            colTypes.add("double");
            colFormat.add("%.5f");
        }
        if (modelCategory == ModelCategory.Binomial) {
            colHeaders.add("Validation Lift");
            colTypes.add("double");
            colFormat.add("%.5f");
        }
        if (isClassifier) {
            colHeaders.add("Validation Classification Error");
            colTypes.add("double");
            colFormat.add("%.5f");
        }
        if (modelCategory == ModelCategory.AutoEncoder) {
            colHeaders.add("Validation MSE");
            colTypes.add("double");
            colFormat.add("%.5f");
        }
    }
    // (hasValidation)
    if (hasCrossValidation) {
        colHeaders.add("Cross-Validation RMSE");
        colTypes.add("double");
        colFormat.add("%.5f");
        if (modelCategory == ModelCategory.Regression) {
            colHeaders.add("Cross-Validation Deviance");
            colTypes.add("double");
            colFormat.add("%.5f");
            colHeaders.add("Cross-Validation MAE");
            colTypes.add("double");
            colFormat.add("%.5f");
        }
        if (isClassifier) {
            colHeaders.add("Cross-Validation LogLoss");
            colTypes.add("double");
            colFormat.add("%.5f");
        }
        if (modelCategory == ModelCategory.Binomial) {
            colHeaders.add("Cross-Validation AUC");
            colTypes.add("double");
            colFormat.add("%.5f");
        }
        if (modelCategory == ModelCategory.Binomial) {
            colHeaders.add("Cross-Validation Lift");
            colTypes.add("double");
            colFormat.add("%.5f");
        }
        if (isClassifier) {
            colHeaders.add("Cross-Validation Classification Error");
            colTypes.add("double");
            colFormat.add("%.5f");
        }
        if (modelCategory == ModelCategory.AutoEncoder) {
            colHeaders.add("Cross-Validation MSE");
            colTypes.add("double");
            colFormat.add("%.5f");
        }
    }
    // (hasCrossValidation)
    final int rows = scoringInfos == null ? 0 : scoringInfos.length;
    String[] s = new String[0];
    TwoDimTable table = new TwoDimTable("Scoring History", null, new String[rows], colHeaders.toArray(s), colTypes.toArray(s), colFormat.toArray(s), "");
    int row = 0;
    if (null == scoringInfos)
        return table;
    for (ScoringInfo si : scoringInfos) {
        int col = 0;
        assert (row < table.getRowDim());
        assert (col < table.getColDim());
        DateTimeFormatter fmt = DateTimeFormat.forPattern("yyyy-MM-dd HH:mm:ss");
        table.set(row, col++, fmt.print(si.time_stamp_ms));
        table.set(row, col++, PrettyPrint.msecs(si.total_training_time_ms, true));
        if (hasSamples) {
            //      Log.info("1st speed: (samples: " + si.training_samples + ", total_run_time: " + si.total_training_time_ms + ", total_scoring_time: " + si.total_scoring_time_ms + ", total_setup_time: " + si.total_setup_time_ms + ")");
            float speed = (float) (((HasSamples) si).training_samples() / ((1. + si.total_training_time_ms - si.total_scoring_time_ms - si.total_setup_time_ms) / 1e3));
            assert (speed >= 0) : "Speed should not be negative! " + speed + " = (float)(" + ((HasSamples) si).training_samples() + "/((" + si.total_training_time_ms + "-" + si.total_scoring_time_ms + "-" + si.total_setup_time_ms + ")/1e3)";
            table.set(row, col++, si.total_training_time_ms == 0 ? null : (speed > 10 ? String.format("%d", (int) speed) : String.format("%g", speed)) + " obs/sec");
        }
        if (hasEpochs)
            table.set(row, col++, ((HasEpochs) si).epoch_counter());
        if (hasIterations)
            table.set(row, col++, ((HasIterations) si).iterations());
        if (hasSamples)
            table.set(row, col++, ((HasSamples) si).training_samples());
        table.set(row, col++, si.scored_train != null ? si.scored_train._rmse : Double.NaN);
        if (modelCategory == ModelCategory.Regression) {
            table.set(row, col++, si.scored_train != null ? si.scored_train._mean_residual_deviance : Double.NaN);
        }
        if (modelCategory == ModelCategory.Regression) {
            table.set(row, col++, si.scored_train != null ? si.scored_train._mae : Double.NaN);
        }
        if (isClassifier) {
            table.set(row, col++, si.scored_train != null ? si.scored_train._logloss : Double.NaN);
        }
        if (modelCategory == ModelCategory.Binomial) {
            table.set(row, col++, si.training_AUC != null ? si.training_AUC._auc : Double.NaN);
            table.set(row, col++, si.scored_train != null ? si.scored_train._lift : Double.NaN);
        }
        if (isClassifier) {
            table.set(row, col++, si.scored_train != null ? si.scored_train._classError : Double.NaN);
        }
        if (isAutoencoder) {
            table.set(row, col++, si.scored_train != null ? si.scored_train._mse : Double.NaN);
        }
        if (hasValidation) {
            table.set(row, col++, si.scored_valid != null ? si.scored_valid._rmse : Double.NaN);
            if (modelCategory == ModelCategory.Regression) {
                table.set(row, col++, si.scored_valid != null ? si.scored_valid._mean_residual_deviance : Double.NaN);
            }
            if (modelCategory == ModelCategory.Regression) {
                table.set(row, col++, si.scored_valid != null ? si.scored_valid._mae : Double.NaN);
            }
            if (isClassifier) {
                table.set(row, col++, si.scored_valid != null ? si.scored_valid._logloss : Double.NaN);
            }
            if (modelCategory == ModelCategory.Binomial) {
                table.set(row, col++, si.validation_AUC != null ? si.validation_AUC._auc : Double.NaN);
                table.set(row, col++, si.scored_valid != null ? si.scored_valid._lift : Double.NaN);
            }
            if (isClassifier) {
                table.set(row, col, si.scored_valid != null ? si.scored_valid._classError : Double.NaN);
            }
            if (isAutoencoder) {
                table.set(row, col++, si.scored_valid != null ? si.scored_valid._mse : Double.NaN);
            }
        }
        // hasValidation
        if (hasCrossValidation) {
            table.set(row, col++, si.scored_xval != null ? si.scored_xval._rmse : Double.NaN);
            if (modelCategory == ModelCategory.Regression) {
                table.set(row, col++, si.scored_xval != null ? si.scored_xval._mean_residual_deviance : Double.NaN);
            }
            if (modelCategory == ModelCategory.Regression) {
                table.set(row, col++, si.scored_xval != null ? si.scored_xval._mae : Double.NaN);
            }
            if (isClassifier) {
                table.set(row, col++, si.scored_xval != null ? si.scored_xval._logloss : Double.NaN);
            }
            if (modelCategory == ModelCategory.Binomial) {
                table.set(row, col++, si.validation_AUC != null ? si.validation_AUC._auc : Double.NaN);
                table.set(row, col++, si.scored_xval != null ? si.scored_xval._lift : Double.NaN);
            }
            if (isClassifier) {
                table.set(row, col, si.scored_xval != null ? si.scored_xval._classError : Double.NaN);
            }
            if (isAutoencoder) {
                table.set(row, col++, si.scored_xval != null ? si.scored_xval._mse : Double.NaN);
            }
        }
        // hasCrossValidation
        row++;
    }
    return table;
}

Example 18 with TwoDimTable

use of water.util.TwoDimTable in project h2o-3 by h2oai.

the class GridSchemaV99 method fillFromImpl.

@Override
public GridSchemaV99 fillFromImpl(Grid grid) {
    Key<Model>[] gridModelKeys = grid.getModelKeys();
    // Return only keys which are referencing to existing objects in DKV
    // However, here is still implicit race, since we are sending
    // keys to client, but referenced models can be deleted in meantime
    // Hence, client has to be responsible for handling this situation
    // - call getModel and check for null model
    // pre-allocate
    List<Key<Model>> modelKeys = new ArrayList<>(gridModelKeys.length);
    for (Key k : gridModelKeys) {
        if (k != null && DKV.get(k) != null) {
            modelKeys.add(k);
        }
    }
    // Default sort order -- TODO: Outsource
    if (sort_by == null && modelKeys.size() > 0 && modelKeys.get(0) != null) {
        Model m = DKV.getGet(modelKeys.get(0));
        if (m != null && m.isSupervised()) {
            if (m._output.nclasses() > 1) {
                sort_by = "logloss";
                decreasing = false;
            } else {
                sort_by = "residual_deviance";
                decreasing = false;
            }
        }
    }
    // If not, show all possible metrics
    if (modelKeys.size() > 0 && sort_by != null) {
        Set<String> possibleMetrics = ModelMetrics.getAllowedMetrics(modelKeys.get(0));
        if (!possibleMetrics.contains(sort_by.toLowerCase())) {
            throw new H2OIllegalArgumentException("Invalid argument for sort_by specified. Must be one of: " + Arrays.toString(possibleMetrics.toArray(new String[0])));
        }
    }
    // Are we sorting by model metrics?
    if (null != sort_by && !sort_by.isEmpty()) {
        // sort the model keys
        modelKeys = ModelMetrics.sortModelsByMetric(sort_by, decreasing, modelKeys);
        // fill the metrics arrays
        training_metrics = new ModelMetricsBaseV3[modelKeys.size()];
        validation_metrics = new ModelMetricsBaseV3[modelKeys.size()];
        cross_validation_metrics = new ModelMetricsBaseV3[modelKeys.size()];
        cross_validation_metrics_summary = new TwoDimTableV3[modelKeys.size()];
        for (int i = 0; i < modelKeys.size(); i++) {
            Model m = DKV.getGet(modelKeys.get(i));
            if (m != null) {
                Model.Output o = m._output;
                if (null != o._training_metrics)
                    training_metrics[i] = (ModelMetricsBaseV3) SchemaServer.schema(3, o._training_metrics).fillFromImpl(o._training_metrics);
                if (null != o._validation_metrics)
                    validation_metrics[i] = (ModelMetricsBaseV3) SchemaServer.schema(3, o._validation_metrics).fillFromImpl(o._validation_metrics);
                if (null != o._cross_validation_metrics)
                    cross_validation_metrics[i] = (ModelMetricsBaseV3) SchemaServer.schema(3, o._cross_validation_metrics).fillFromImpl(o._cross_validation_metrics);
                if (o._cross_validation_metrics_summary != null)
                    cross_validation_metrics_summary[i] = new TwoDimTableV3(o._cross_validation_metrics_summary);
            }
        }
    }
    KeyV3.ModelKeyV3[] modelIds = new KeyV3.ModelKeyV3[modelKeys.size()];
    Key<Model>[] keys = new Key[modelKeys.size()];
    for (int i = 0; i < modelIds.length; i++) {
        modelIds[i] = new KeyV3.ModelKeyV3(modelKeys.get(i));
        keys[i] = modelIds[i].key();
    }
    grid_id = new KeyV3.GridKeyV3(grid._key);
    model_ids = modelIds;
    hyper_names = grid.getHyperNames();
    failed_params = toModelParametersSchema(grid.getFailedParameters());
    failure_details = grid.getFailureDetails();
    failure_stack_traces = grid.getFailureStackTraces();
    failed_raw_params = grid.getFailedRawParameters();
    TwoDimTable t = grid.createSummaryTable(keys, sort_by, decreasing);
    if (t != null)
        summary_table = new TwoDimTableV3().fillFromImpl(t);
    TwoDimTable h = grid.createScoringHistoryTable();
    if (h != null)
        scoring_history = new TwoDimTableV3().fillFromImpl(h);
    return this;
}

Example 19 with TwoDimTable

use of water.util.TwoDimTable in project h2o-3 by h2oai.

the class TwoDimTableTest method run3.

@Test
public void run3() {
    TwoDimTable table = new TwoDimTable("My foo bar table", "desc", new String[] { "First row", "R2", "Row #3", "Last row is here:" }, new String[] { "DoubleValue", "S2", "My Terrible Percent Value" }, new String[] { "double", "string", "double" }, new String[] { "%f", "%s", "%f" }, "", new String[][] { new String[] { null, "One", null }, new String[] { null, null, null }, new String[] { null, "Three", null }, new String[] { null, "FooBar", null } }, new double[][] { new double[] { 1.123, emptyDouble, 3200034.00001 }, new double[] { 123.34, emptyDouble, 1.0 }, new double[] { emptyDouble, emptyDouble, 3234.00001 }, new double[] { 3.33420923423423, emptyDouble, 3.40234234 } });
    String ts = table.toString();
    assertTrue(ts.length() > 0);
    Log.info(ts);
    String json = new TwoDimTableV3().fillFromImpl(table).toJsonString();
    Log.info(json);
}

Example 20 with TwoDimTable

use of water.util.TwoDimTable in project h2o-3 by h2oai.

the class DeepWaterModelInfo method createSummaryTable.

/**
   * Create a summary table
   * @return TwoDimTable with the summary of the model
   */
TwoDimTable createSummaryTable() {
    TwoDimTable table = new TwoDimTable("Status of Deep Learning Model", (get_params()._network == null ? ("MLP: " + Arrays.toString(get_params()._hidden)) : get_params()._network.toString()) + ", " + PrettyPrint.bytes(size()) + ", " + (!get_params()._autoencoder ? ("predicting " + get_params()._response_column + ", ") : "") + (get_params()._autoencoder ? "auto-encoder" : _classification ? (_classes + "-class classification") : "regression") + ", " + String.format("%,d", get_processed_global()) + " training samples, " + "mini-batch size " + String.format("%,d", get_params()._mini_batch_size), //rows
    new String[1], new String[] { "Input Neurons", "Rate", "Momentum" }, new String[] { "int", "double", "double" }, new String[] { "%d", "%5f", "%5f" }, "");
    table.set(0, 0, _dataInfo != null ? _dataInfo.fullN() : _width * _height * _channels);
    table.set(0, 1, get_params().learningRate(get_processed_global()));
    table.set(0, 2, get_params().momentum(get_processed_global()));
    summaryTable = table;
    return summaryTable;
}