cavis/libnd4j/include/ops/declarable/generic/linalg/qr.cpp

90 lines
4.3 KiB
C++
Raw Normal View History

Shugeo qr (#153) * Added qr op implementation. Initial version. * Fixed doc for qr op. Signed-off-by: shugeo <sgazeos@gmail.com> * Implementation of QR decomposition. CPU platform version. * Added a pair of tests for qr op testing. Signed-off-by: shugeo <sgazeos@gmail.com> * QR implementation. Signed-off-by: shugeo <sgazeos@gmail.com> * Corrected norm using. * Properly calculated intermediate results with QR decomposition. * Another step to implement QR algorithm by householder. * Cpu implementatio for QR decomposition. The first working edition. * Corrected test to QR decomposition. * Added tad multithreading with QR implementation. * Finished cpu implementation for QR decomposition helpers. * Refactored tests and improved multithreading. * Refactored QR cpu implementation and update cuda implementation helpers. * Cuda QR helper implementation. The first working edition. * Eliminated waste prints. * Restore multithreading with cuda implementation. * Ops names corrected * Refactored qr op helpers to optimize. Signed-off-by: shugeo <sgazeos@gmail.com> * Eliminated waste manual ticking. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored memory allocation to avoid waste memory usage. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored matrixMinor method both for cuda and cpu platforms. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored method of vmul to use raw buffers instead type conversion. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored temporary array of matricies. Signed-off-by: shugeo <sgazeos@gmail.com> Co-authored-by: Alexander Stoyakin <alexander.stoyakin@gmail.com> Co-authored-by: raver119 <raver119@gmail.com>
2020-01-22 11:59:36 +01:00
/*******************************************************************************
* Copyright (c) 2019-2020 Konduit K.K.
*
* 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
******************************************************************************/
//
// Created by GS <sgazeos@gmail.com> at 12/20/2019
//
#include <system/op_boilerplate.h>
Shugeo qr (#153) * Added qr op implementation. Initial version. * Fixed doc for qr op. Signed-off-by: shugeo <sgazeos@gmail.com> * Implementation of QR decomposition. CPU platform version. * Added a pair of tests for qr op testing. Signed-off-by: shugeo <sgazeos@gmail.com> * QR implementation. Signed-off-by: shugeo <sgazeos@gmail.com> * Corrected norm using. * Properly calculated intermediate results with QR decomposition. * Another step to implement QR algorithm by householder. * Cpu implementatio for QR decomposition. The first working edition. * Corrected test to QR decomposition. * Added tad multithreading with QR implementation. * Finished cpu implementation for QR decomposition helpers. * Refactored tests and improved multithreading. * Refactored QR cpu implementation and update cuda implementation helpers. * Cuda QR helper implementation. The first working edition. * Eliminated waste prints. * Restore multithreading with cuda implementation. * Ops names corrected * Refactored qr op helpers to optimize. Signed-off-by: shugeo <sgazeos@gmail.com> * Eliminated waste manual ticking. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored memory allocation to avoid waste memory usage. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored matrixMinor method both for cuda and cpu platforms. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored method of vmul to use raw buffers instead type conversion. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored temporary array of matricies. Signed-off-by: shugeo <sgazeos@gmail.com> Co-authored-by: Alexander Stoyakin <alexander.stoyakin@gmail.com> Co-authored-by: raver119 <raver119@gmail.com>
2020-01-22 11:59:36 +01:00
#include <ops/declarable/CustomOperations.h>
#include <ops/declarable/helpers/qr.h>
#if NOT_EXCLUDED(OP_qr)
namespace sd {
Shugeo qr (#153) * Added qr op implementation. Initial version. * Fixed doc for qr op. Signed-off-by: shugeo <sgazeos@gmail.com> * Implementation of QR decomposition. CPU platform version. * Added a pair of tests for qr op testing. Signed-off-by: shugeo <sgazeos@gmail.com> * QR implementation. Signed-off-by: shugeo <sgazeos@gmail.com> * Corrected norm using. * Properly calculated intermediate results with QR decomposition. * Another step to implement QR algorithm by householder. * Cpu implementatio for QR decomposition. The first working edition. * Corrected test to QR decomposition. * Added tad multithreading with QR implementation. * Finished cpu implementation for QR decomposition helpers. * Refactored tests and improved multithreading. * Refactored QR cpu implementation and update cuda implementation helpers. * Cuda QR helper implementation. The first working edition. * Eliminated waste prints. * Restore multithreading with cuda implementation. * Ops names corrected * Refactored qr op helpers to optimize. Signed-off-by: shugeo <sgazeos@gmail.com> * Eliminated waste manual ticking. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored memory allocation to avoid waste memory usage. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored matrixMinor method both for cuda and cpu platforms. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored method of vmul to use raw buffers instead type conversion. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored temporary array of matricies. Signed-off-by: shugeo <sgazeos@gmail.com> Co-authored-by: Alexander Stoyakin <alexander.stoyakin@gmail.com> Co-authored-by: raver119 <raver119@gmail.com>
2020-01-22 11:59:36 +01:00
namespace ops {
CUSTOM_OP_IMPL(qr, 1, 2, false, 0, 0) {
auto input = INPUT_VARIABLE(0);
auto outputQ = OUTPUT_VARIABLE(0);
auto outputR = OUTPUT_VARIABLE(1);
auto fullMatricies = false;
if (block.getBArguments()->size())
fullMatricies = B_ARG(0);
REQUIRE_TRUE(input->rankOf() >=2, 0, "qr: The rank of input array should not be less than 2, but %i is given", input->rankOf());
REQUIRE_TRUE((fullMatricies && outputQ->sizeAt(-1) == input->sizeAt(-2)) || (!fullMatricies && outputQ->isSameShape(input)), 0, "qr: The last dimmensions should be equal to result Q, but %i and %i are given", outputQ->sizeAt(-1), input->sizeAt(-2));
REQUIRE_TRUE((fullMatricies && outputR->sizeAt(-1) == input->sizeAt(-1)) || (!fullMatricies && outputR->sizeAt(-1) == outputR->sizeAt(-2)), 0, "qr: The last dimmensions should be equal to result R, but %i and %i are given", outputR->sizeAt(-1), input->sizeAt(-1));
Shugeo solve ls (#203) * lstsq op. Initial commit. Signed-off-by: shugeo <sgazeos@gmail.com> * Least squares linear problem solve op (lstsq). Cpu draft implementation. Signed-off-by: shugeo <sgazeos@gmail.com> * Fixed shape routine and tests. Signed-off-by: shugeo <sgazeos@gmail.com> * Added test for lstsq op. Signed-off-by: shugeo <sgazeos@gmail.com> * Rectification for lstsq op implementation. Signed-off-by: shugeo <sgazeos@gmail.com> * Corrected test to avoid numerical inconsistensy. Signed-off-by: shugeo <sgazeos@gmail.com> * Added prints for check computing. Signed-off-by: shugeo <sgazeos@gmail.com> * Corrected tests to use evalueate facility instead. Signed-off-by: shugeo <sgazeos@gmail.com> * CPU implementation of MatrixSolveLs op and tests. Signed-off-by: shugeo <sgazeos@gmail.com> * Added cuda implementation for helpers with lstsq op. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored tests for lstsq op. Signed-off-by: shugeo <sgazeos@gmail.com> * Added processing for empty inputs. Signed-off-by: shugeo <sgazeos@gmail.com> * Merged tests. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored lstsq op for fast case. Signed-off-by: shugeo <sgazeos@gmail.com> * Fixed test. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored lstsq op. Signed-off-by: shugeo <sgazeos@gmail.com> * Fixed some issues with solve. Signed-off-by: shugeo <sgazeos@gmail.com> * Fixed lstsq op to avoid erros. Signed-off-by: shugeo <sgazeos@gmail.com> * Added kernel for giagonal factor Signed-off-by: shugeo <sgazeos@gmail.com> * lstsq wrapper and triangular_solve fixed * Added proper processing empty inputs and test. Signed-off-by: shugeo <sgazeos@gmail.com> * SequenceMask test * Build fixed * Added proper processing of empty inputs with solve op. Signed-off-by: shugeo <sgazeos@gmail.com> * Mapping added * Added check of input shapes with solve op. Signed-off-by: shugeo <sgazeos@gmail.com> * Added a couple of tests for lstsq op and minor changes with cuda helper for one.' Signed-off-by: shugeo <sgazeos@gmail.com> * Tests on * Refactored test for lstsq op. Signed-off-by: shugeo <sgazeos@gmail.com> * Fixed test * Added another approach for lstsq op aka solve_ls. Signed-off-by: shugeo <sgazeos@gmail.com> * Finished cpu part for solve_ls op helpers. * Added helper for low triangular matrix inversion. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored alternate solve_ls cpu implementation. Signed-off-by: shugeo <sgazeos@gmail.com> * Removed alternate approach for solve_ls op. Added multithreading with matrix inversion. Signed-off-by: shugeo <sgazeos@gmail.com> * Assert fixed * Refactored multithreading for inverse matricies. Signed-off-by: shugeo <sgazeos@gmail.com> Co-authored-by: Alexander Stoyakin <alexander.stoyakin@gmail.com>
2020-02-28 09:37:26 +01:00
if (!input->isEmpty() && !outputQ->isEmpty() && !outputR->isEmpty())
helpers::qr(block.launchContext(), input, outputQ, outputR, fullMatricies);
Shugeo qr (#153) * Added qr op implementation. Initial version. * Fixed doc for qr op. Signed-off-by: shugeo <sgazeos@gmail.com> * Implementation of QR decomposition. CPU platform version. * Added a pair of tests for qr op testing. Signed-off-by: shugeo <sgazeos@gmail.com> * QR implementation. Signed-off-by: shugeo <sgazeos@gmail.com> * Corrected norm using. * Properly calculated intermediate results with QR decomposition. * Another step to implement QR algorithm by householder. * Cpu implementatio for QR decomposition. The first working edition. * Corrected test to QR decomposition. * Added tad multithreading with QR implementation. * Finished cpu implementation for QR decomposition helpers. * Refactored tests and improved multithreading. * Refactored QR cpu implementation and update cuda implementation helpers. * Cuda QR helper implementation. The first working edition. * Eliminated waste prints. * Restore multithreading with cuda implementation. * Ops names corrected * Refactored qr op helpers to optimize. Signed-off-by: shugeo <sgazeos@gmail.com> * Eliminated waste manual ticking. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored memory allocation to avoid waste memory usage. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored matrixMinor method both for cuda and cpu platforms. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored method of vmul to use raw buffers instead type conversion. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored temporary array of matricies. Signed-off-by: shugeo <sgazeos@gmail.com> Co-authored-by: Alexander Stoyakin <alexander.stoyakin@gmail.com> Co-authored-by: raver119 <raver119@gmail.com>
2020-01-22 11:59:36 +01:00
return Status::OK();
}
DECLARE_SHAPE_FN(qr) {
auto inShape = inputShape->at(0);
Nd4jLong* shapeQ;
Nd4jLong* shapeR;
int targetRank = shape::rank(inShape); // last two dimensions will be reduced to scalar
auto fullMatricies = false;
if (block.getBArguments()->size())
fullMatricies = B_ARG(0);
auto shape = ShapeUtils::shapeAsVector(inShape);
if (!fullMatricies) { // outputs are: Q is MxN and R is NxN
shape[targetRank - 1] = shape::sizeAt(inShape, -1);
shape[targetRank - 2] = shape[targetRank - 1];
shapeQ = ConstantShapeHelper::getInstance()->createShapeInfo(ArrayOptions::dataType(inShape),
shape::order(inShape), targetRank,
shape::shapeOf(inShape));
shapeR = ConstantShapeHelper::getInstance()->createShapeInfo(ArrayOptions::dataType(inShape),
shape::order(inShape), shape);
}
else {// otherwise outputs are Q is MxM and R is MxN with zero filled rows
shape[targetRank - 1] = shape::sizeAt(inShape, -2);
shape[targetRank - 2] = shape[targetRank - 1];
shapeR = ConstantShapeHelper::getInstance()->createShapeInfo(ArrayOptions::dataType(inShape),
shape::order(inShape), targetRank,
shape::shapeOf(inShape));
shapeQ = ConstantShapeHelper::getInstance()->createShapeInfo(ArrayOptions::dataType(inShape),
shape::order(inShape), shape);
}
return SHAPELIST(shapeQ, shapeR);
}
DECLARE_TYPES(qr) {
getOpDescriptor()
->setAllowedInputTypes({ALL_FLOATS})
->setAllowedOutputTypes({ALL_FLOATS});
}
}
}
#endif