cavis/libnd4j/include/ops/declarable/helpers/cuda/updaterAdaMax.cu

143 lines
7.2 KiB
Plaintext
Raw Normal View History

Learning updaters for gradient (#335) * libnd4j raw implementation of sgd upader Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j some corrections and simple test added Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j some corrections after discussion Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j integrate applyScalar Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j raw implementation of rmsPropUpdater on cpu Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j fix operations declaration Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j rmsPropUpdater added, test cases for sgd, etc Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j fixed several typos Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j some fixes and improvements for rmsPropUpdater based on Java tests Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j fixed cuda implementation, update tests and corrected behavior according java tests Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j adaGrad updater added Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j one minor fix for ada grad Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j several more fixes for ada_grad Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j nesterovs updater added Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j fixed nesterovs updater behavior, several typos and rename file Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j one minor typo Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j ada max updater added Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j fixed several typos in adaMax updater Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j fixed several typos in adaMaxUpdater Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j several fixes for adaMax, added Adam Updater Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j adaDeltaUpdater added, minor fixes for adamUpdater Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j several fixes for adaDeltaUpdater Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j nadamUpdater added Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j one more correction for nadam updater Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j several fixes for nadam updater and added amsGradUpdater Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j several typos fixed in amsGradUpdater Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j some corrections and added f order support rmsProp updater Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j added support of f order for all updaters and modify tests for testing in place Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j fixed issues for updates when not in place mode used, added tests for f order Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j added input shape checks Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j some corrections for different cases handling Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j some code clean up and optimize per request Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j updaters refactoring after review Signed-off-by: Oleg <oleg.semeniv@gmail.com> * SgdUpdater wrapper Signed-off-by: raver119 <raver119@gmail.com> * first test Signed-off-by: raver119 <raver119@gmail.com> * RmsPropUpdater added Signed-off-by: raver119 <raver119@gmail.com> * NadamUpdater + NesterovsUpdater Signed-off-by: raver119 <raver119@gmail.com> * AmsGradUpdater Signed-off-by: raver119 <raver119@gmail.com> * AdamUpdater added Signed-off-by: raver119 <raver119@gmail.com> * AdaGradUpdater + AdaDeltaUpdater + AdaMaxUpdater Signed-off-by: raver119 <raver119@gmail.com> * AdaGradUpdater test added Signed-off-by: raver119 <raver119@gmail.com> * libnd4j remove input parameters parsing through NDArray, split implementation of helpers to separate files, added some rename, etc Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j next step to split operations implementation into separate files Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j merge master and minor corrections Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j revert some changes of split implementation Signed-off-by: Oleg <oleg.semeniv@gmail.com> * libnd4j forgot to add header file Signed-off-by: Oleg <oleg.semeniv@gmail.com> * public default constructors Signed-off-by: raver119 <raver119@gmail.com> * ImportClassMapping updated Signed-off-by: raver119 <raver119@gmail.com> Co-authored-by: raver119 <raver119@gmail.com>
2020-03-23 05:28:31 +01:00
/*******************************************************************************
* Copyright (c) 2019 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
******************************************************************************/
//
// @author Oleh Semeniv (oleg.semeniv@gmail.com)
//
#include <system/op_boilerplate.h>
#include <ops/declarable/helpers/updatersHelpers.h>
#include <helpers/PointersManager.h>
#include <math/platformmath.h>
#include <math/templatemath.h>
namespace sd {
namespace ops {
namespace helpers {
///////////////////////////////////////////////////////////////////
template<typename T>
__global__ void adaMaxUpdaterCuda(const void* vx, const Nd4jLong* xShapeInfo, const void* vinv, const Nd4jLong* invShapeInfo,
const void* vinm, const Nd4jLong* inmShapeInfo, void* vz, const Nd4jLong* zShapeInfo,
void* vstV, const Nd4jLong* stvShapeInfo, void* vstM, const Nd4jLong* stmShapeInfo,
const T lr, const T beta1, const T beta2, const T epsilon, const T iteration) {
const auto grad = reinterpret_cast<const T*>(vx);
const auto initU = reinterpret_cast<const T*>(vinv);
const auto initM = reinterpret_cast<const T*>(vinm);
auto up = reinterpret_cast<T*>(vz);
auto stU = reinterpret_cast<T*>(vstV);
auto stM = reinterpret_cast<T*>(vstM);
__shared__ Nd4jLong xLen;
__shared__ T beta1T, epsilonT;
__shared__ bool bEWS, bOrdering, bXZsame, bXInUSame, bXStUSame, bXInMSame, bXStMSame;
if (threadIdx.x == 0) {
xLen = shape::length(xShapeInfo);
beta1T = sd::math::nd4j_pow<T,T,T>(beta1, (iteration + 1) );
epsilonT = lr / (1.0 - beta1T);
if (sd::math::nd4j_isnan(epsilonT) || 0 == epsilonT || sd::math::nd4j_isinf(epsilonT))
epsilonT = epsilon;
bEWS = 1 == shape::elementWiseStride(xShapeInfo) && 1 == shape::elementWiseStride(zShapeInfo) &&
1 == shape::elementWiseStride(stmShapeInfo) && 1 == shape::elementWiseStride(inmShapeInfo) &&
1 == shape::elementWiseStride(stvShapeInfo) && 1 == shape::elementWiseStride(invShapeInfo);
bOrdering = shape::order(xShapeInfo) == shape::order(zShapeInfo) && shape::order(xShapeInfo) == shape::order(stmShapeInfo) &&
shape::order(xShapeInfo) == shape::order(inmShapeInfo) && shape::order(xShapeInfo) == shape::order(invShapeInfo) &&
shape::order(xShapeInfo) == shape::order(stvShapeInfo);
bXZsame = shape::haveSameShapeAndStrides(xShapeInfo, zShapeInfo);
bXInUSame = shape::haveSameShapeAndStrides(xShapeInfo, invShapeInfo);
bXStUSame = shape::haveSameShapeAndStrides(xShapeInfo, stvShapeInfo);
bXInMSame = shape::haveSameShapeAndStrides(xShapeInfo, inmShapeInfo);
bXStMSame = shape::haveSameShapeAndStrides(xShapeInfo, stmShapeInfo);
}
__syncthreads();
int coords[MAX_RANK];
for (Nd4jLong i = blockIdx.x * blockDim.x + threadIdx.x; i < xLen; i += gridDim.x * blockDim.x) {
auto xOffset = i, zOffset = i, initMOffset = i, initUOffset = i, stMOffset = i, stUOffset = i;
if (!bEWS || !bOrdering) {
shape::index2coords(i, xShapeInfo, coords);
xOffset = shape::getOffset(xShapeInfo, coords);
zOffset = bXZsame ? xOffset : shape::getOffset(zShapeInfo, coords);
initUOffset = bXInUSame ? xOffset : shape::getOffset(invShapeInfo, coords);
stUOffset = bXStUSame ? xOffset : shape::getOffset(stvShapeInfo, coords);
initMOffset = bXInMSame ? xOffset : shape::getOffset(inmShapeInfo, coords);
stMOffset = bXStMSame ? xOffset : shape::getOffset(stmShapeInfo, coords);
}
//m = B_1 * m + (1-B_1)*grad
stM[stMOffset] = beta1 * initM[initMOffset] + grad[xOffset] * (1 - beta1);
//u = max(B_2 * u, |grad|)
stU[stUOffset] = sd::math::nd4j_max( (beta2* initU[initUOffset]), sd::math::nd4j_abs(grad[xOffset])) + 1e-32;
up[zOffset] = (stM[stMOffset] * epsilonT) / stU[stUOffset];
}
}
///////////////////////////////////////////////////////////////////
template<typename T>
linkage void adaMaxUpdaterCudaLauncher(const int blocksPerGrid, const int threadsPerBlock, const cudaStream_t* stream, const void* vx, const Nd4jLong* xShapeInfo,
const void* vinv, const Nd4jLong* invShapeInfo, const void* vinm, const Nd4jLong* inmShapeInfo,
void* vz, const Nd4jLong* zShapeInfo, void* vstV, const Nd4jLong* stvShapeInfo,
void* vstM, const Nd4jLong* stmShapeInfo, const double dLr,
const double dBeta1, const double dBeta2, const double dEpsilon, const int nIteration) {
const T lr = static_cast<T>(dLr);
const T beta1 = static_cast<T>(dBeta1);
const T beta2 = static_cast<T>(dBeta2);
const T epsilon = static_cast<T>(dEpsilon);
const T iteration = static_cast<T>(nIteration);
adaMaxUpdaterCuda<T> << <blocksPerGrid, threadsPerBlock, 256, * stream >> > (vx, xShapeInfo, vinv, invShapeInfo, vinm, inmShapeInfo, vz,
zShapeInfo, vstV, stvShapeInfo, vstM, stmShapeInfo, lr, beta1, beta2, epsilon, iteration);
}
///////////////////////////////////////////////////////////////////
void updaterAdaMax(sd::LaunchContext* context, const NDArray& gradient, const NDArray& initStateU, const NDArray& initStateM,
NDArray& update, NDArray& stateU, NDArray& stateM, const double dLr, const double dBeta1,
const double dBeta2, const double dEpsilon, const int nIteration) {
PointersManager manager(context, "adaMaxUpdater");
const int threadsPerBlock = MAX_NUM_THREADS / 4;
const int blocksPerGrid = (gradient.lengthOf() + threadsPerBlock - 1) / threadsPerBlock;
NDArray::prepareSpecialUse({ &update, &stateU, &stateM }, { &gradient, &initStateU, &initStateM });
BUILD_SINGLE_SELECTOR(gradient.dataType(), adaMaxUpdaterCudaLauncher, (blocksPerGrid, threadsPerBlock, context->getCudaStream(),
gradient.getSpecialBuffer(), gradient.getSpecialShapeInfo(), initStateU.getSpecialBuffer(),
initStateU.getSpecialShapeInfo(), initStateM.getSpecialBuffer(), initStateM.getSpecialShapeInfo(),
update.getSpecialBuffer(), update.getSpecialShapeInfo(), stateU.getSpecialBuffer(),
stateU.getSpecialShapeInfo(), stateM.getSpecialBuffer(), stateM.getSpecialShapeInfo(),
dLr, dBeta1, dBeta2, dEpsilon, nIteration ), FLOAT_TYPES);
NDArray::registerSpecialUse({ &update, &stateU, &stateM }, { &gradient, &initStateU, &initStateM });
manager.synchronize();
}
}
}
}