Examples with OperationTypes org.apache.sysml.lops.ConvolutionTransform.OperationTypes used on opensource projects

Example 1 with OperationTypes

use of org.apache.sysml.lops.ConvolutionTransform.OperationTypes in project incubator-systemml by apache.

the class ConvolutionOp method constructConvolutionLops.

public Lop constructConvolutionLops(ExecType et, ArrayList<Hop> inputs) throws HopsException, LopsException {
    if (inputs.size() != getNumExpectedInputs())
        throw new HopsException("Incorrect number of inputs for " + op.name());
    Lop in = null;
    Lop in2 = null;
    ArrayList<Hop> inputs1 = inputs;
    int k = OptimizerUtils.getConstrainedNumThreads(_maxNumThreads);
    OperationTypes lopOp = HopsConv2Lops.get(op);
    // For other backends, this operators is not necessary as it reduces an additional relu operator.
    if (OptimizerUtils.ALLOW_OPERATOR_FUSION && et == ExecType.CP && op == ConvOp.MAX_POOLING && isInputReLU(inputs.get(0))) {
        in = inputs.get(0).getInput().get(0).constructLops();
        lopOp = OperationTypes.RELU_MAX_POOLING;
    } else if (OptimizerUtils.ALLOW_OPERATOR_FUSION && et == ExecType.CP && op == ConvOp.MAX_POOLING_BACKWARD && isInputReLU(inputs.get(0))) {
        in = inputs.get(0).getInput().get(0).constructLops();
        lopOp = OperationTypes.RELU_MAX_POOLING_BACKWARD;
    } else if (OptimizerUtils.ALLOW_OPERATOR_FUSION && op == ConvOp.BIAS_ADD && isInputConv2d(inputs.get(0))) {
        lopOp = OperationTypes.DIRECT_CONV2D_BIAS_ADD;
        // the first lop is image 
        in = inputs.get(0).getInput().get(0).constructLops();
        // the second lop is bias
        in2 = inputs.get(1).constructLops();
        // Use the inputs from conv2d rather than bias_add
        inputs1 = inputs.get(0).getInput();
    } else {
        in = inputs.get(0).constructLops();
    }
    //		// TODO: Inserting reblock requires knowing columns apriori
    //		ConvolutionTransform transform1 = new ConvolutionTransform(addReblockIfNecessary(et, lopOp, in), lopOp, getDataType(), getValueType(), et, k);
    //		setReblockedOutputDimension(et, transform1);
    ConvolutionTransform transform1 = new ConvolutionTransform(in, lopOp, getDataType(), getValueType(), et, k);
    setOutputDimensions(transform1);
    setLineNumbers(transform1);
    in.addOutput(transform1);
    if (in2 != null) {
        transform1.addInput(in2);
        in2.addOutput(transform1);
    }
    // filter_shape1, filter_shape2, filter_shape3, filter_shape4
    for (int i = 1; i < inputs1.size(); i++) {
        Lop ltmp = inputs1.get(i).constructLops();
        transform1.addInput(ltmp);
        ltmp.addOutput(transform1);
    }
    //force order of added lops
    transform1.setLevel();
    return transform1;
}