org.nd4j.autodiff.validation.OpTestCase Maven / Gradle / Ivy
/*******************************************************************************
* Copyright (c) 2015-2018 Skymind, Inc.
*
* This program and the accompanying materials are made available under the
* terms of the Apache License, Version 2.0 which is available at
* https://www.apache.org/licenses/LICENSE-2.0.
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
* License for the specific language governing permissions and limitations
* under the License.
*
* SPDX-License-Identifier: Apache-2.0
******************************************************************************/
package org.nd4j.autodiff.validation;
import lombok.Data;
import lombok.NonNull;
import lombok.experimental.Accessors;
import org.nd4j.autodiff.validation.functions.EqualityFn;
import org.nd4j.autodiff.validation.functions.RelErrorFn;
import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.api.ops.DynamicCustomOp;
import org.nd4j.linalg.api.shape.LongShapeDescriptor;
import org.nd4j.common.function.Function;
import java.util.HashMap;
import java.util.LinkedHashMap;
import java.util.Map;
/**
* Validate the output - and shape function - of a single operation.
*
* Used with {@link OpValidation}
*
* @author Alex Black
*/
@Data
@Accessors(fluent = true)
public class OpTestCase {
private final DynamicCustomOp op;
private Map> testFns = new LinkedHashMap<>();
private Map expShapes = new HashMap<>();
public OpTestCase(@NonNull DynamicCustomOp op) {
this.op = op;
}
/**
* Validate the op output using INDArray.equals(INDArray)
*
* @param outputNum Number of the output
* @param expected Expected INDArray
*/
public OpTestCase expectedOutput(int outputNum, INDArray expected) {
testFns.put(outputNum, new EqualityFn(expected));
expShapes.put(outputNum, expected.shapeDescriptor());
return this;
}
/**
* Validate the output for a single variable using element-wise relative error:
* relError = abs(x-y)/(abs(x)+abs(y)), with x=y=0 case defined to be 0.0.
* Also has a minimum absolute error condition, which must be satisfied for the relative error failure to be considered
* legitimate
*
* @param outputNum output number
* @param expected Expected INDArray
* @param maxRelError Maximum allowable relative error
* @param minAbsError Minimum absolute error for a failure to be considered legitimate
*/
public OpTestCase expectedOutputRelError(int outputNum, @NonNull INDArray expected, double maxRelError, double minAbsError) {
testFns.put(outputNum, new RelErrorFn(expected, maxRelError, minAbsError));
expShapes.put(outputNum, expected.shapeDescriptor());
return this;
}
/**
* @param outputNum Output number to check
* @param expShape Expected shape for the output
* @param validationFn Function to use to validate the correctness of the specific Op. Should return null
* if validation passes, or an error message if the op validation fails
*/
public OpTestCase expectedOutput(int outputNum, @NonNull LongShapeDescriptor expShape, @NonNull Function validationFn) {
testFns.put(outputNum, validationFn);
expShapes.put(outputNum, expShape);
return this;
}
}