Examples with FloatTensor com.alibaba.alink.common.linalg.tensor.FloatTensor used on opensource projects

Example 1 with FloatTensor

use of com.alibaba.alink.common.linalg.tensor.FloatTensor in project Alink by alibaba.

the class TFExampleConversionV2 method javaToFeature.

/**
 * convert java object to tensorflow feature.
 *
 * @param dt  java object data type.
 * @param val given java object.
 * @return tensorflow feature.
 */
public static Feature javaToFeature(DataTypesV2 dt, Object val) {
    Feature.Builder featureBuilder = Feature.newBuilder();
    FloatList.Builder floatListBuilder = FloatList.newBuilder();
    Int64List.Builder int64ListBuilder = Int64List.newBuilder();
    // When dt is TENSOR, find the exact type first.
    if (DataTypesV2.TENSOR.equals(dt)) {
        if (val instanceof FloatTensor) {
            dt = DataTypesV2.FLOAT_TENSOR;
        } else if (val instanceof DoubleTensor) {
            dt = DataTypesV2.DOUBLE_TENSOR;
        } else if (val instanceof IntTensor) {
            dt = DataTypesV2.INT_TENSOR;
        } else if (val instanceof LongTensor) {
            dt = DataTypesV2.LONG_TENSOR;
        } else if (val instanceof BoolTensor) {
            dt = DataTypesV2.BOOLEAN_TENSOR;
        } else if (val instanceof UByteTensor) {
            dt = DataTypesV2.UBYTE_TENSOR;
        } else if (val instanceof StringTensor) {
            dt = DataTypesV2.STRING_TENSOR;
        } else if (val instanceof ByteTensor) {
            dt = DataTypesV2.BYTE_TENSOR;
        }
    }
    switch(dt) {
        case FLOAT_16:
        case FLOAT_32:
        case FLOAT_64:
            {
                floatListBuilder.addValue((Float) val);
                featureBuilder.setFloatList(floatListBuilder);
                break;
            }
        case INT_8:
        case INT_16:
        case INT_32:
        case INT_64:
        case UINT_8:
        case UINT_16:
        case UINT_32:
        case UINT_64:
            {
                int64ListBuilder.addValue(castAsLong(val));
                featureBuilder.setInt64List(int64ListBuilder);
                break;
            }
        case STRING:
            {
                BytesList.Builder bb = BytesList.newBuilder();
                bb.addValue(castAsBytes(val));
                featureBuilder.setBytesList(bb);
                break;
            }
        case FLOAT_TENSOR:
            {
                FloatTensor floatTensor = (FloatTensor) val;
                long size = floatTensor.size();
                floatTensor = floatTensor.reshape(new Shape(size));
                for (long i = 0; i < size; i += 1) {
                    floatListBuilder.addValue(floatTensor.getFloat(i));
                }
                featureBuilder.setFloatList(floatListBuilder);
                break;
            }
        case DOUBLE_TENSOR:
            {
                DoubleTensor doubleTensor = (DoubleTensor) val;
                long size = doubleTensor.size();
                doubleTensor = doubleTensor.reshape(new Shape(size));
                for (long i = 0; i < size; i += 1) {
                    floatListBuilder.addValue((float) doubleTensor.getDouble(i));
                }
                featureBuilder.setFloatList(floatListBuilder);
                break;
            }
        case INT_TENSOR:
            {
                IntTensor intTensor = (IntTensor) val;
                long size = intTensor.size();
                intTensor = intTensor.reshape(new Shape(size));
                for (long i = 0; i < size; i += 1) {
                    int64ListBuilder.addValue(intTensor.getInt(i));
                }
                featureBuilder.setInt64List(int64ListBuilder);
                break;
            }
        case LONG_TENSOR:
            {
                LongTensor longTensor = (LongTensor) val;
                long size = longTensor.size();
                longTensor = longTensor.reshape(new Shape(size));
                for (long i = 0; i < size; i += 1) {
                    int64ListBuilder.addValue(longTensor.getLong(i));
                }
                featureBuilder.setInt64List(int64ListBuilder);
                break;
            }
        case BOOLEAN_TENSOR:
            {
                BoolTensor boolTensor = (BoolTensor) val;
                long size = boolTensor.size();
                boolTensor = boolTensor.reshape(new Shape(size));
                for (long i = 0; i < size; i += 1) {
                    int64ListBuilder.addValue(boolTensor.getBoolean(i) ? 1 : 0);
                }
                featureBuilder.setInt64List(int64ListBuilder);
                break;
            }
        case UBYTE_TENSOR:
            {
                UByteTensor ubyteTensor = (UByteTensor) val;
                long size = ubyteTensor.size();
                ubyteTensor = ubyteTensor.reshape(new Shape(size));
                for (long i = 0; i < size; i += 1) {
                    int64ListBuilder.addValue(ubyteTensor.getUByte(i));
                }
                featureBuilder.setInt64List(int64ListBuilder);
                break;
            }
        case STRING_TENSOR:
            {
                StringTensor stringTensor = (StringTensor) val;
                long size = stringTensor.size();
                stringTensor = stringTensor.reshape(new Shape(size));
                BytesList.Builder bb = BytesList.newBuilder();
                for (long i = 0; i < size; i += 1) {
                    bb.addValue(castAsBytes(stringTensor.getString(i)));
                }
                featureBuilder.setBytesList(bb);
                break;
            }
        case BYTE_TENSOR:
        default:
            throw new RuntimeException("Unsupported data type for TF");
    }
    return featureBuilder.build();
}

Example 2 with FloatTensor

use of com.alibaba.alink.common.linalg.tensor.FloatTensor in project Alink by alibaba.

the class DeepARModelMapper method predictMultiVar.

@Override
protected Tuple2<Vector[], String> predictMultiVar(Timestamp[] historyTimes, Vector[] historyVals, int predictNum) {
    Timestamp[] predictTimes = TimeSeriesMapper.getPredictTimes(historyTimes, predictNum);
    int window = historyVals.length;
    int series = 0;
    DenseVector[] vectors = new DenseVector[historyVals.length];
    for (int i = 0; i < window; ++i) {
        vectors[i] = VectorUtil.getDenseVector(historyVals[i]);
        if (vectors[i] == null) {
            throw new IllegalArgumentException("history values should not be null.");
        }
        series = vectors[i].size();
    }
    FloatTensor[][] tensors = new FloatTensor[series][window];
    for (int i = 0; i < series; ++i) {
        tensors[i][0] = Tensor.cat(new FloatTensor[] { new FloatTensor(new float[] { 0.0f }), DeepARFeaturesGenerator.generateFromFrequency(calendar.get(), unit, historyTimes[0]) }, -1, null);
        for (int j = 1; j < window; ++j) {
            tensors[i][j] = Tensor.cat(new FloatTensor[] { new FloatTensor(new float[] { (float) vectors[j - 1].get(i) }), DeepARFeaturesGenerator.generateFromFrequency(calendar.get(), unit, historyTimes[j]) }, -1, null);
        }
    }
    FloatTensor[] batch = new FloatTensor[series];
    for (int i = 0; i < series; ++i) {
        batch[i] = Tensor.stack(tensors[i], 0, null);
    }
    Vector[] result = new Vector[predictNum];
    Row[] sigmas = new Row[predictNum];
    for (int i = 0; i < predictNum; ++i) {
        result[i] = new DenseVector(series);
        sigmas[i] = Row.of(new DenseVector(series));
    }
    for (int i = 0; i < series; ++i) {
        float mu = (float) historyVals[window - 1].get(i);
        FloatTensor v = new FloatTensor(new float[] { 0.0f, 0.0f });
        int nonZero = 0;
        for (int j = 0; j < window; ++j) {
            float cell = batch[i].getFloat(j, 0);
            if (cell != 0) {
                nonZero += 1;
            }
            v.setFloat(v.getFloat(0) + cell, 0);
        }
        if (mu != 0) {
            nonZero += 1;
            v.setFloat(v.getFloat(0) + mu, 0);
        }
        if (nonZero == 0) {
            continue;
        }
        v.setFloat(v.getFloat(0) / nonZero + 1.0f, 0);
        for (int j = 0; j < window; ++j) {
            batch[i].setFloat(batch[i].getFloat(j, 0) / v.getFloat(0), j, 0);
        }
        mu = mu / v.getFloat(0);
        for (int j = 0; j < predictNum; ++j) {
            batch[i] = Tensor.cat(new FloatTensor[] { batch[i], Tensor.stack(new FloatTensor[] { Tensor.cat(new FloatTensor[] { new FloatTensor(new float[] { mu }), DeepARFeaturesGenerator.generateFromFrequency(calendar.get(), unit, predictTimes[j]) }, -1, null) }, 0, null) }, 0, null);
            FloatTensor pred;
            try {
                pred = (FloatTensor) tfTableModelPredictModelMapper.map(Row.of(batch[i])).getField(0);
            } catch (Exception e) {
                return Tuple2.of(null, null);
            }
            mu = pred.getFloat(window + j, 0);
            float sigma = pred.getFloat(window + j, 1);
            result[j].set(i, mu * v.getFloat(0) + v.getFloat(1));
            ((Vector) (sigmas[j].getField(0))).set(i, sigma * v.getFloat(0));
        }
    }
    return Tuple2.of(result, new MTable(Arrays.asList(sigmas), new String[] { "sigma" }, new TypeInformation<?>[] { VectorTypes.DENSE_VECTOR }).toString());
}

Example 3 with FloatTensor

use of com.alibaba.alink.common.linalg.tensor.FloatTensor in project Alink by alibaba.

the class DeepARModelMapper method predictSingleVar.

@Override
protected Tuple2<double[], String> predictSingleVar(Timestamp[] historyTimes, double[] historyVals, int predictNum) {
    Timestamp[] predictTimes = TimeSeriesMapper.getPredictTimes(historyTimes, predictNum);
    int window = historyVals.length;
    FloatTensor[] tensors = new FloatTensor[window];
    // fill the first z with zero
    tensors[0] = Tensor.cat(new FloatTensor[] { new FloatTensor(new float[] { 0.0f }), DeepARFeaturesGenerator.generateFromFrequency(calendar.get(), unit, historyTimes[0]) }, -1, null);
    // others
    for (int i = 1; i < window; ++i) {
        tensors[i] = Tensor.cat(new FloatTensor[] { new FloatTensor(new float[] { (float) historyVals[i - 1] }), DeepARFeaturesGenerator.generateFromFrequency(calendar.get(), unit, historyTimes[i]) }, -1, null);
    }
    FloatTensor batch = Tensor.stack(tensors, 0, null);
    // initialize mu
    float mu = (float) historyVals[window - 1];
    // calculate v
    FloatTensor v = new FloatTensor(new float[] { 0.0f, 0.0f });
    int nonZero = 0;
    for (int i = 0; i < window; ++i) {
        float cell = batch.getFloat(i, 0);
        if (cell != 0) {
            nonZero += 1;
        }
        v.setFloat(v.getFloat(0) + cell, 0);
    }
    if (mu != 0) {
        nonZero += 1;
        v.setFloat(v.getFloat(0) + mu, 0);
    }
    if (nonZero == 0) {
        double[] result = new double[predictNum];
        Row[] sigmas = new Row[predictNum];
        Arrays.fill(result, 0.0);
        Arrays.fill(sigmas, Row.of(0));
        return Tuple2.of(result, new MTable(Arrays.asList(sigmas), new String[] { "sigma" }, new TypeInformation<?>[] { Types.DOUBLE }).toString());
    }
    v.setFloat(v.getFloat(0) / nonZero + 1.0f, 0);
    // normalize with v
    for (int i = 0; i < window; ++i) {
        batch.setFloat(batch.getFloat(i, 0) / v.getFloat(0), i, 0);
    }
    mu = mu / v.getFloat(0);
    // result initialize.
    double[] result = new double[predictNum];
    Row[] sigmas = new Row[predictNum];
    Arrays.fill(result, 0.0);
    for (int i = 0; i < predictNum; ++i) {
        sigmas[i] = Row.of(0.0);
    }
    // prediction
    for (int j = 0; j < predictNum; ++j) {
        batch = Tensor.cat(new FloatTensor[] { batch, Tensor.stack(new FloatTensor[] { Tensor.cat(new FloatTensor[] { new FloatTensor(new float[] { mu }), DeepARFeaturesGenerator.generateFromFrequency(calendar.get(), unit, predictTimes[j]) }, -1, null) }, 0, null) }, 0, null);
        FloatTensor pred;
        try {
            pred = (FloatTensor) tfTableModelPredictModelMapper.map(Row.of(batch)).getField(0);
        } catch (Exception e) {
            return Tuple2.of(null, null);
        }
        mu = pred.getFloat(window + j, 0);
        float sigma = pred.getFloat(window + j, 1);
        result[j] = mu * v.getFloat(0) + v.getFloat(1);
        sigmas[j].setField(0, sigma * v.getFloat(0));
    }
    return Tuple2.of(result, new MTable(Arrays.asList(sigmas), new String[] { "sigma" }, new TypeInformation<?>[] { Types.DOUBLE }).toString());
}

Example 4 with FloatTensor

use of com.alibaba.alink.common.linalg.tensor.FloatTensor in project Alink by alibaba.

the class LSTNetModelMapper method predictMultiVar.

@Override
protected Tuple2<Vector[], String> predictMultiVar(Timestamp[] historyTimes, Vector[] historyVals, int predictNum) {
    Tuple2<Timestamp[], FloatTensor> t = toTensor(historyTimes, historyVals);
    FloatTensor pred = null;
    try {
        pred = (FloatTensor) tfTableModelPredictModelMapper.map(Row.of(t.f1)).getField(0);
    } catch (Exception ex) {
    // pass
    }
    return pred == null ? Tuple2.of(null, null) : Tuple2.of(new Vector[] { DoubleTensor.of(pred).toVector() }, null);
}

Example 5 with FloatTensor

use of com.alibaba.alink.common.linalg.tensor.FloatTensor in project Alink by alibaba.

the class FlattenMTableStreamTest method linkFrom.

@Test
public void linkFrom() throws Exception {
    List<Row> rows = new ArrayList<>();
    rows.add(Row.of(1, "2", 0, null, new SparseVector(3, new int[] { 1 }, new double[] { 2.0 }), new FloatTensor(new float[] { 3.0f })));
    rows.add(Row.of(null, "2", 0, new DenseVector(new double[] { 0.0, 1.0 }), new SparseVector(4, new int[] { 2 }, new double[] { 3.0 }), new FloatTensor(new float[] { 3.0f })));
    rows.add(Row.of(null, "2", 0, new DenseVector(new double[] { 0.1, 1.0 }), new SparseVector(4, new int[] { 2 }, new double[] { 3.0 }), new FloatTensor(new float[] { 3.0f })));
    String schemaStr = "col0 int, col1 string, label int" + ", d_vec DENSE_VECTOR" + ", s_vec SPARSE_VECTOR" + ", tensor FLOAT_TENSOR";
    MTable mTable = new MTable(rows, schemaStr);
    List<Row> table = new ArrayList<>();
    table.add(Row.of("id", mTable.toString()));
    StreamOperator<?> op = new MemSourceStreamOp(table, new String[] { "id", "mTable" });
    StreamOperator<?> res = op.link(new FlattenMTableStreamOp().setSchemaStr(schemaStr).setSelectedCol("mTable").setReservedCols("id"));
    CollectSinkStreamOp sop = res.link(new CollectSinkStreamOp());
    StreamOperator.execute();
    List<Row> list = sop.getAndRemoveValues();
    for (Row row : list) {
        Assert.assertEquals(row.getField(0), "id");
    }
}