cavis/libnd4j/include/ops/declarable/helpers/cpu/triu.cpp

59 lines
2.0 KiB
C++
Raw Normal View History

2021-02-01 13:31:45 +01:00
/* ******************************************************************************
*
Backpropagation implementation of mergemax, mergeadd and mergeavg ops (#343) * libnd4j: first step of merge_max implementation Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j fixed typos Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j some corrections for mergeMaxBp Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j some minor corrections Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j test added for mergemax_bp Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j fixed several problems tests added, check with gradCheck Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j remove duplicated tests Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j split implementation of transforms ops into separate file implementation Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j code clean up, added mergeavg_bp and mergeadd_bp, need testing Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j merge master, fixed typos and added tests Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j some minor fixes Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j added helper for mergeAddBp operation, this permits to skip nullify Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j file renaming changes and cuda some corrections, need some additional corrections Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j some additional corrections for merge ops Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j more corrections per request for cuda more proper usage Signed-off-by: Oleg <oleg.semeniv@gmail.com>
2020-03-25 06:40:30 +01:00
*
* 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.
*
2021-02-01 13:31:45 +01:00
* See the NOTICE file distributed with this work for additional
* information regarding copyright ownership.
Backpropagation implementation of mergemax, mergeadd and mergeavg ops (#343) * libnd4j: first step of merge_max implementation Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j fixed typos Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j some corrections for mergeMaxBp Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j some minor corrections Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j test added for mergemax_bp Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j fixed several problems tests added, check with gradCheck Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j remove duplicated tests Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j split implementation of transforms ops into separate file implementation Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j code clean up, added mergeavg_bp and mergeadd_bp, need testing Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j merge master, fixed typos and added tests Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j some minor fixes Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j added helper for mergeAddBp operation, this permits to skip nullify Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j file renaming changes and cuda some corrections, need some additional corrections Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j some additional corrections for merge ops Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j more corrections per request for cuda more proper usage Signed-off-by: Oleg <oleg.semeniv@gmail.com>
2020-03-25 06:40:30 +01:00
* 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
******************************************************************************/
//
// @author Yurii Shyrma (iuriish@yahoo.com), created on 20.04.2018
//
#include <ops/declarable/helpers/transforms.h>
#include <helpers/Loops.h>
namespace sd {
namespace ops {
namespace helpers {
//////////////////////////////////////////////////////////////////////////
template <typename T>
static void triuBP_(sd::LaunchContext * context, const NDArray& input, const NDArray& gradO, NDArray& gradI, const int diagonal) {
auto dOdI = NDArray(&gradO); // dO/dI
const_cast<NDArray&>(input).fillAsTriangular<T>(0, diagonal, dOdI.sizeAt(-1), dOdI, 'b');
int dLen = dOdI.lengthOf();
auto func = PRAGMA_THREADS_FOR {
for (auto i = start; i < stop; i++) {
if (dOdI.t<T>(i) != static_cast<T>(0.f))
Shyrma sqrtm (#429) * - start working on implementation of sqrtm op Signed-off-by: Yurii <iuriish@yahoo.com> * - improving householder procedure Signed-off-by: Yurii <iuriish@yahoo.com> * - further polishing householder stuff Signed-off-by: Yurii <iuriish@yahoo.com> * - polishing hh pivoting qr procedure Signed-off-by: Yurii <iuriish@yahoo.com> * - polishing BiDiagonalUp procedure Signed-off-by: Yurii <iuriish@yahoo.com> * - polishing householder sequence class Signed-off-by: Yurii <iuriish@yahoo.com> * - polishing jacobi svd class Signed-off-by: Yurii <iuriish@yahoo.com> * - polishing svd stuff 1 Signed-off-by: Yurii <iuriish@yahoo.com> * - polishing svd stuff 2 Signed-off-by: Yurii <iuriish@yahoo.com> * - implementation and testing class which performs Hessenberg decomposition of square matrix Signed-off-by: Yurii <iuriish@yahoo.com> * - add static method to JacobiSVD class which makes the continuous Givens rotation generation algorithm Signed-off-by: Yurii <iuriish@yahoo.com> * - implementation and testing auxiliary methods of Schur decomp class Signed-off-by: Yurii <iuriish@yahoo.com> * some references here and there Signed-off-by: raver119 <raver119@gmail.com> * - trying figure out difference between eigen and our Schur alg Signed-off-by: Yurii <iuriish@yahoo.com> * - testing fixing bugs in Schur decomposition op Signed-off-by: Yurii <iuriish@yahoo.com> * - start to implement class which performs calculation of eigen values and vectors Signed-off-by: Yurii <iuriish@yahoo.com> * - add to EigenValsAndVecs method which calculates complex eigen vectors Signed-off-by: Yurii <iuriish@yahoo.com> * - testing and fixing bugs in EigenValsAndVecs class Signed-off-by: Yurii <iuriish@yahoo.com> * - implementation and testing triangularSolver class Signed-off-by: Yurii <iuriish@yahoo.com> * Added a 2D routine for triangular systems solve. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored triangularSolve2D routine and tests. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored another test for triangularSolve2D. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored test for triangularSolve for vector-bar case. Signed-off-by: shugeo <sgazeos@gmail.com> * Refactored triangularSolve2D routine and tests. Signed-off-by: shugeo <sgazeos@gmail.com> * - implementation of FullPivLU class Signed-off-by: Yurii <iuriish@yahoo.com> * - fix bugs in FullPivLU::solve method Signed-off-by: Yurii <iuriish@yahoo.com> * - correct permutation vector in FullPivLU::solve Signed-off-by: Yurii <iuriish@yahoo.com> * - correct include headers Signed-off-by: Yurii <iuriish@yahoo.com> * - implementation of Sqrtm class Signed-off-by: Yurii <iuriish@yahoo.com> * - testing and fixing bugs in Sqrtm class Signed-off-by: Yurii <iuriish@yahoo.com> * - include sqrtm classes to cuda folder, investigate in what places synchronization doesn't work Signed-off-by: Yurii <iuriish@yahoo.com> * Added implementation for cuda triangularSolve2D and also refactored triangularSolve2D for cpu. Signed-off-by: shugeo <sgazeos@gmail.com> * Eliminated waste implementations. Signed-off-by: shugeo <sgazeos@gmail.com> * - make offset calculation faster in t<> methods Signed-off-by: Yurii <iuriish@yahoo.com> * - rename refference T& NDArray::t<> method Signed-off-by: Yurii <iuriish@yahoo.com> * - further work on cuda sqrtm Signed-off-by: Yurii <iuriish@yahoo.com> * - provide correct synchronization to device in Sqrtm class Signed-off-by: Yurii <iuriish@yahoo.com> * - add tests for sqrtm op Signed-off-by: Yurii <iuriish@yahoo.com> * - correct fails which appeared while testing on jenkins Signed-off-by: Yurii <iuriish@yahoo.com> * - trying to find out mistake in svd::deflation method Signed-off-by: Yurii <iuriish@yahoo.com> * Revert "- trying to find out mistake in svd::deflation method" This reverts commit 19d37baddbc509028e4bc67bc932fe7449becdb6. * Revert "- trying to find out mistake in svd::deflation method" This reverts commit 19d37baddbc509028e4bc67bc932fe7449becdb6. Signed-off-by: Yurii <iuriish@yahoo.com> * - change call semantic of r<> and t<> methods Signed-off-by: Yurii <iuriish@yahoo.com> * - ged rid of ambiguity in * operator overloads for windows buikd Signed-off-by: Yurii <iuriish@yahoo.com> * - get rid of ambiguity in * operator overloads for windows build 2 Signed-off-by: Yurii <iuriish@yahoo.com> * - get rid of ambiguity in * operator overloads for windows build 3 Signed-off-by: Yurii <iuriish@yahoo.com> * - resolve conflicts with master Signed-off-by: Yurii <iuriish@yahoo.com> * cmakelists updated Signed-off-by: raver119@gmail.com <raver119@gmail.com> * - minor fix in merge cpu helper - make use of reference getter Signed-off-by: Yurii <iuriish@yahoo.com> Co-authored-by: raver119 <raver119@gmail.com> Co-authored-by: shugeo <sgazeos@gmail.com>
2020-05-14 17:06:13 +02:00
dOdI.r<T>(i) = static_cast<T>(1.f);
Backpropagation implementation of mergemax, mergeadd and mergeavg ops (#343) * libnd4j: first step of merge_max implementation Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j fixed typos Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j some corrections for mergeMaxBp Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j some minor corrections Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j test added for mergemax_bp Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j fixed several problems tests added, check with gradCheck Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j remove duplicated tests Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j split implementation of transforms ops into separate file implementation Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j code clean up, added mergeavg_bp and mergeadd_bp, need testing Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j merge master, fixed typos and added tests Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j some minor fixes Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j added helper for mergeAddBp operation, this permits to skip nullify Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j file renaming changes and cuda some corrections, need some additional corrections Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j some additional corrections for merge ops Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j more corrections per request for cuda more proper usage Signed-off-by: Oleg <oleg.semeniv@gmail.com>
2020-03-25 06:40:30 +01:00
}
};
samediff::Threads::parallel_for(func, 0, dLen);
// FIXME: !!!
gradI.assign(dOdI * gradO); // chain rule: dLoss/dI = dO/dI * dLoss/dO
}
void triuBP(sd::LaunchContext * context, const NDArray& input, const NDArray& gradO, NDArray& gradI, const int diagonal) {
BUILD_SINGLE_SELECTOR(gradO.dataType(), triuBP_, (context, input, gradO, gradI, diagonal), LIBND4J_TYPES);
}
}
}
}