org.nd4j.linalg.api.ops.impl.shape.Concat Maven / Gradle / Ivy

Show more of this group Show more artifacts with this name
Show all versions of nd4j-api Show documentation
There is a newer version: 1.0.0-M2.1
package org.nd4j.linalg.api.ops.impl.shape;

import lombok.extern.slf4j.Slf4j;
import lombok.val;
import onnx.OnnxProto3;
import org.nd4j.autodiff.samediff.SDVariable;
import org.nd4j.autodiff.samediff.SameDiff;
import org.nd4j.imports.NoOpNameFoundException;
import org.nd4j.imports.descriptors.properties.PropertyMapping;
import org.nd4j.linalg.api.ops.DynamicCustomOp;
import org.nd4j.linalg.api.ops.Op;
import org.nd4j.linalg.exception.ND4JIllegalStateException;
import org.tensorflow.framework.AttrValue;
import org.tensorflow.framework.GraphDef;
import org.tensorflow.framework.NodeDef;

import java.util.HashMap;
import java.util.LinkedHashMap;
import java.util.Map;

@Slf4j
public class Concat extends DynamicCustomOp {
    private int concatDimension;

    public Concat(){

    }

    public Concat(SameDiff sameDiff, int concatDimension, SDVariable... inputs){
        super(null, sameDiff, inputs);
        addIArgument(concatDimension);
    }

    @Override
    public String opName() {
        return "concat";
    }


    @Override
    public void resolvePropertiesFromSameDiffBeforeExecution() {
        val propertiesToResolve = sameDiff.propertiesToResolveForFunction(this);
        if(!propertiesToResolve.isEmpty()) {
            val varName = propertiesToResolve.get(0);
            val var = sameDiff.getVariable(varName);
            if(var == null) {
                throw new ND4JIllegalStateException("No variable found with name " +varName);
            }
            else if(var.getArr() == null) {
                throw new ND4JIllegalStateException("Array with variable name " + varName + " unset!");
            }

            concatDimension = var.getArr().getInt(0);
            addIArgument(concatDimension);
        }

        //don't pass both iArg and last axis down to libnd4j
        if(inputArguments().length == args().length) {
            val inputArgs = inputArguments();
            removeInputArgument(inputArgs[inputArguments().length - 1]);
        }
    }

    @Override
    public void assertValidForExecution() {
        val descriptor = getDescriptor();
        if(descriptor == null)
            throw new NoOpNameFoundException("No descriptor found for op name " + opName());


        if(descriptor.getNumInputs() > 0 && numInputArguments() < 2)
            throw new ND4JIllegalStateException("Op failure for " + opName() + " Number of inputs is invalid for execution. Specified " + numInputArguments() + " but should be " + descriptor.getNumInputs());

        if(descriptor.getNumOutputs() > 0 && numOutputArguments() != descriptor.getNumOutputs())
            throw new ND4JIllegalStateException("Op failure for " + opName() + " Number of outputs is invalid for execution. Specified " + numOutputArguments() + " but should be " + descriptor.getNumOutputs());

        //< 0 means dynamic size
        if(descriptor.getNumIArgs() >= 0 && numIArguments() != descriptor.getNumIArgs())
            throw new ND4JIllegalStateException("Op failure for " + opName() + " Number of integer arguments is invalid for execution. Specified " + numIArguments() + " but should be " + descriptor.getNumIArgs());

        if(descriptor.getNumTArgs() >= 0 && numTArguments() != descriptor.getNumTArgs())
            throw new ND4JIllegalStateException("Op failure for " + opName() + " Number of inputs is invalid for execution. Specified " + numTArguments() + " but should be " + descriptor.getNumTArgs());

    }

    @Override
    public Map> mappingsForFunction() {
        Map> ret = new HashMap<>();

        Map concatMap = new HashMap<>();
        val concatDimProps = PropertyMapping.builder()
                .tfInputPosition(0)
                .onnxAttrName("axis")
                .build();
        concatMap.put("concatDimension",concatDimProps);


        Map concatV2Map = new HashMap<>();
        val concat2DimProps = PropertyMapping.builder()
                //lalst position
                .tfInputPosition(-1)
                .onnxAttrName("axis")
                .build();
        concatV2Map.put("concatDimension",concat2DimProps);

        //note that onnx is already covered here
        ret.put(tensorflowNames()[0],concatMap);
        ret.put(tensorflowNames()[1],concatV2Map);


        return ret;
    }

    @Override
    public void initFromTensorFlow(NodeDef nodeDef, SameDiff initWith, Map attributesForNode, GraphDef graph) {
        int concatDimension = -1;
        String input = null;
        for(int i = 0; i < nodeDef.getInputCount(); i++) {
            if(nodeDef.getInput(i).contains("/concat_dim")) {
                input = nodeDef.getInput(i);
                break;
            }
        }

        //older versions may specify a concat_dim, usually it's the last argument
        if(input == null) {
            input = nodeDef.getInput(nodeDef.getInputCount() - 1);
        }

        val variable = initWith.getVariable(input);
        // concat dimension is only possible
        if (variable != null && variable.getArr() == null) {
            sameDiff.addPropertyToResolve(this, input);

        } else if (variable != null) {
            val arr = variable.getArr();
            if (arr.length() == 1) {
                concatDimension = arr.getInt(0);
            }

            this.concatDimension = concatDimension;
            addIArgument(this.concatDimension);
            log.debug("Concat dimension: {}", concatDimension);

        }

        //don't pass both iArg and last axis down to libnd4j
        if(inputArguments().length == nodeDef.getInputCount()) {
            val inputArgs = inputArguments();
            removeInputArgument(inputArgs[inputArguments().length - 1]);
        }

        sameDiff.removeArgFromFunction(input,this);
    }

    @Override
    public Map propertiesForFunction() {
        Map ret = new LinkedHashMap<>();
        ret.put("concatDimension",concatDimension);
        return ret;
    }


    @Override
    public void initFromOnnx(OnnxProto3.NodeProto node, SameDiff initWith, Map attributesForNode, OnnxProto3.GraphProto graph) {
        super.initFromOnnx(node, initWith, attributesForNode, graph);
    }

    @Override
    public String onnxName() {
        return "Concat";
    }

    @Override
    public String tensorflowName() {
        return "Concat";
    }


    @Override
    public String[] tensorflowNames() {
        return new String[]  {"Concat","ConcatV2"};
    }

    @Override
    public Op.Type opType() {
        return Op.Type.CUSTOM;
    }
}