cavis/libnd4j/blas/cpu/NDArray.cpp

454 lines
18 KiB
C++
Raw Normal View History

2019-06-06 14:21:15 +02:00
/*******************************************************************************
* 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
******************************************************************************/
#ifndef NDARRAY_CPP
#define NDARRAY_CPP
#include "../NDArray.h"
#include "../NDArrayFactory.h"
#include "NativeOpExecutioner.h"
#include <BroadcastPairwiseConverter.h>
#include <memory/Workspace.h>
#include <memory/MemoryRegistrator.h>
#include <ops.h>
#include <ops/gemm.h>
#include <pointercast.h>
#include <stdexcept>
#include <memory>
#include <helpers/logger.h>
#include <loops/pairwise_transform.h>
#include <loops/transform_same.h>
#include <loops/random.h>
#include <loops/broadcasting.h>
#include <indexing/NDIndex.h>
#include <indexing/IndicesList.h>
#include <helpers/ShapeUtils.h>
#include <sstream>
#include <helpers/ArrayUtils.h>
#include <MmulHelper.h>
#include <helpers/threshold.h>
#include <exceptions/datatype_exception.h>
#include <exceptions/allocation_exception.h>
#include <helpers/ConstantTadHelper.h>
#include <NDArray.hpp>
namespace nd4j {
////////////////////////////////////////////////////////////////////////
void* NDArray::platformBuffer() { return buffer(); }
void* NDArray::getPlatformBuffer() const { return getBuffer(); }
Nd4jLong* NDArray::getPlatformShapeInfo() const { return getShapeInfo(); }
Nd4jLong* NDArray::platformShapeInfo() { return shapeInfo(); }
void NDArray::syncToDevice() const { }
void NDArray::syncToHost() const { }
void NDArray::tickWriteHost() const { }
void NDArray::tickWriteDevice() const { }
void NDArray::tickReadHost() const { }
void NDArray::tickReadDevice() const { }
void NDArray::tickBothActual() const { }
Dev branch merge: dev_20190606 (#7904) * correct logsoftmax looss (#2) * Small SameDiff listener fix (#4) * Various fixes (#6) * #7839 Fix for asXMatrix and tests * #7866 EmbeddingSequenceLayer dtype fix + test * #7856 SameDiff save/load stream methods * #7859 RegressionEvaluation rank 4 fix + tests + axis configuration * EvaluationBinary 3d/4d * More evaluation 3d/4d tests * #7847 Evaluation empty checks * Small test ifx * #7848 Fix median edge case * Improve DL4J samediff layer tests * [WIP] FastText wrapper implemented (#8) * FastText implemented * Some fixes * Fix shapes for wordsNearest * Validation of input vectors * Fixes * Fixed test * Thread tagged * Some tweaks * setContextClassLoader for DeallocatorServiceThread * Numpy format tests (#1) * Various fixes (#11) * #7852 SameDiff gather fix * #7892 SameDiff placeholder to constant conversion * #7890 validate input rank for MLN/CG init methods * Fix broken permute shape calculation * Permute and gather fixes * Tests * #7850 LogSumExp fix + test * Handful of test fixes * Empty arrays with non-scalar shapes (#10) * minor rearrangements for lambdas * empty tensors with non-scalar shapes * numpy empty tensors with non-scalar shapes * few more empty tweaks * Small fixes * conv3d signature update * micro fix in batchnorm mkldnn * Import fixes * Fix * MKL-DNN update * Small fill fix * fill with empty input + test * Fixes * Small error improvement * Fix * one special test * couple of fixes for lstm * Rewrite TFGraphMapper.getNDArrayFromTensor to be maintainable and less error prone * Fixes * FP16 * Unsigned * BFloat16 * Fill op - empty tweaks * - couple of fixes for empty arrays construction - stack updated * strided slice fix * one transform test * provide method for reducing shapeInfo in case of input array is empty * Fixed reduceAlongDimensions to use empty input properly. * couple of broadcast tests * couple of tests broadcast tests + tweak to make them pass * add check of non-empty to methods producing sub-arrays * Fixed reshapeC with zeros in shape. * complete empty check in reduce_... legacy ops * Concat and cumsum/prod * Tweak to empty shape inference on import * add empty check to the rest of reduce legacy ops * one more test * correct typo in evalReduceShapeInfoEmpty * Added tests for reduce_* ops to tests with zero shapes. * few more tests for empty reductions * Fixed strided_slice op with empty case and tests. * one more empty reduction test * Fixed strided_slice test. * add empty check to NDArray::reshapei * infOrMax * empty min/max with infinity tests * made unstack working correctly with empty arrays * few IndexReduce tests + tweaks for empty shapes * add test for empty concat * few tests fixed * Validation fix for reductions on empty shapes * Reverse fix * Reduction shape calc fixes * SameDiff.generateOutputVariable: don't use shape function to determine number of outputs * Range fix * - NDArray constructor updated for scalars/empty arrays - few tests fixed * More fixes * Empty creator fixes * concat fix * concat fix * TF import tests: allow 'both all NaN' and 'both all inf' to pass * Slice, zero fraction, and reshape fixes * transpose, gather * Zero fraction * scalar cast fix * Empty reduction axis support * few more tests fixed * Fixed input checks conforming with TF for concat op and tests. * few tests fixed * matmul scalar shape fix * Fixed checkout for data type and scalarity with concat to allow non-empty scalars with vector concats. * broadcast bool fix * few more tests * few more tests * correct evalReduceShapeInfoEmpty * argmax/argmin + tests * one more empty edge case + one more test * argmax/argmin/realdiv_bp tweaks * empty reshape test + fix * Helper fixes * Small fixes * Gather test fix * Gather test fix * Small fixes * reduce scalar zero values * scalar mean workaround * Remove debug code * along dim mean workaround * one more test * - equalsTo() tweak for empty arrays - one more test * broadcast tweaks
2019-06-15 13:34:34 +02:00
bool NDArray::isActualOnHostSide() const { return true; }
bool NDArray::isActualOnDeviceSide() const { return true; }
2019-06-06 14:21:15 +02:00
void NDArray::makeBothBuffersActual() const { }
////////////////////////////////////////////////////////////////////////
template <typename T>
void NDArray::fillAsTriangular(const float val, int lower, int upper, const char direction, NDArray* target) {
if (isS())
throw std::runtime_error("NDArray::fillArrayAsTriangular: you can't use this method on String array!");
if(target == nullptr)
target = this;
if(!isSameShape(target) && !(rankOf() == 1 && target->rankOf() == 2 && sizeAt(0) == target->sizeAt(0) && sizeAt(0) == target->sizeAt(1)))
throw std::string("NDArray::fillArrayAsTriangular method: wrong shape of target array !");
if (direction == 'u')
lower = -target->sizeAt(-2);
else if (direction == 'l')
upper = target->sizeAt(-1);
const T value = static_cast<T>(val);
const auto x = reinterpret_cast<const T*>(getBuffer());
auto z = reinterpret_cast<T*>(target->getBuffer());
const int xRank = rankOf();
const int zRank = target->rankOf();
const auto zLen = target->lengthOf();
const bool areSameOffsets = shape::haveSameShapeAndStrides(getShapeInfo(), target->getShapeInfo());
std::vector<Nd4jLong> coords(zRank);
PRAGMA_OMP_PARALLEL_FOR_ARGS(if(zLen > Environment::getInstance()->elementwiseThreshold()) firstprivate(coords))
for (Nd4jLong i = 0; i < zLen; ++i) {
shape::index2coords(zRank, target->shapeOf(), i, zLen, coords.data());
const auto zOffset = shape::getOffset(0, target->shapeOf(), target->stridesOf(), coords.data(), zRank);
// if( (row + upper < col) || (row + lower > col) )
if((coords[zRank - 2] + upper < coords[zRank - 1]) || (coords[zRank - 2] + lower > coords[zRank - 1]))
z[zOffset] = value;
else if(this != target) { // when this and target are different arrays
if(xRank != zRank)
coords[0] = coords[1];
const auto xOffset = areSameOffsets ? zOffset : shape::getOffset(0, shapeOf(), stridesOf(), coords.data(), xRank);
z[zOffset] = x[xOffset];
}
}
}
BUILD_SINGLE_TEMPLATE(template void NDArray::fillAsTriangular, (const float val, int lower, int upper, const char direction, NDArray* target), LIBND4J_TYPES);
////////////////////////////////////////////////////////////////////////
void NDArray::setIdentity() {
if (isS())
throw std::runtime_error("NDArray::setIdentity: you can't use this method on String array!");
this->nullify();
int rank = rankOf();
auto shape = shapeOf();
auto strides = stridesOf();
int minDim = MAX_INT;
Nd4jLong indices[MAX_RANK];
for(int j = 0; j < rank; ++j)
indices[j] = 1;
Nd4jLong offset = shape::getOffset(0, shape, strides, indices, rank);
for(int i = 0; i < rank; ++i)
if(minDim > shape[i])
minDim = shape[i];
float v = 1.0f;
PRAGMA_OMP_PARALLEL_FOR_ARGS(if(minDim > Environment::getInstance()->elementwiseThreshold()) schedule(guided))
for(int i = 0; i < minDim; ++i)
templatedSet<float>(buffer(), i*offset, this->dataType(), &v);
}
////////////////////////////////////////////////////////////////////////
template <typename T>
static void templatedSwap(void *xBuffer, void *yBuffer, Nd4jLong length) {
auto x = reinterpret_cast<T *>(xBuffer);
auto y = reinterpret_cast<T *>(yBuffer);
PRAGMA_OMP_PARALLEL_FOR_SIMD_ARGS(schedule(static))
for (Nd4jLong i = 0; i < length; ++i) {
2019-06-06 14:21:15 +02:00
auto temp = x[i];
x[i] = y[i];
y[i] = temp;
}
}
BUILD_SINGLE_TEMPLATE(template void templatedSwap, (void *xBuffer, void *yBuffer, Nd4jLong length), LIBND4J_TYPES);
////////////////////////////////////////////////////////////////////////
void NDArray::swapUnsafe(NDArray& other) {
auto xType = this->dataType();
if (xType != other.dataType())
throw std::runtime_error("NDArray::swapUnsage method: both arrays must have the same data type");
if(buffer() == nullptr || other.buffer() == nullptr)
throw std::runtime_error("NDArray::swapUnsafe method: input array should not be empty!");
if(lengthOf() != other.lengthOf())
throw std::runtime_error("NDArray::swapUnsafe method: input arrays should have the same length!");
BUILD_SINGLE_SELECTOR(xType, templatedSwap, (buffer(), other.buffer(), this->lengthOf()), LIBND4J_TYPES);
}
////////////////////////////////////////////////////////////////////////
void NDArray::synchronize(const char* msg) const {
// no-op
}
void NDArray::prepareSpecialUse(const std::initializer_list<const NDArray*>& writeList, const std::initializer_list<const NDArray*>& readList, bool synchronizeWritables) {
2019-06-06 14:21:15 +02:00
// no-op
}
void NDArray::registerSpecialUse(const std::initializer_list<const NDArray*>& writeList, const std::initializer_list<const NDArray*>& readList) {
2019-06-06 14:21:15 +02:00
// no-op
}
void NDArray::preparePrimaryUse(const std::initializer_list<const NDArray*>& writeList, const std::initializer_list<const NDArray*>& readList, bool synchronizeWritables) {
2019-06-06 14:21:15 +02:00
// no-op
}
void NDArray::registerPrimaryUse(const std::initializer_list<const NDArray*>& writeList, const std::initializer_list<const NDArray*>& readList) {
2019-06-06 14:21:15 +02:00
// no-op
}
2019-06-06 14:21:15 +02:00
void NDArray::syncShape() const {
// no-op
}
//////////////////////////////////////////////////////////////////////////
template<typename T>
void NDArray::printCurrentBuffer(const bool host, const char* msg, const int precision) const {
}
////////////////////////////////////////////////////////////////////////
void* NDArray::specialBufferWithOffset(Nd4jLong offset) const {
return nullptr;
}
Merge master to upstream (#7945) * Shugeo strided slice zeros (#14) * Modified strided_slice op to properly work with empty-like shapes. * Fixed test for reduce_mean with empty-like input. * [WIP] Last merge (#15) * correct logsoftmax looss (#2) * Small SameDiff listener fix (#4) * Various fixes (#6) * #7839 Fix for asXMatrix and tests * #7866 EmbeddingSequenceLayer dtype fix + test * #7856 SameDiff save/load stream methods * #7859 RegressionEvaluation rank 4 fix + tests + axis configuration * EvaluationBinary 3d/4d * More evaluation 3d/4d tests * #7847 Evaluation empty checks * Small test ifx * #7848 Fix median edge case * Improve DL4J samediff layer tests * [WIP] FastText wrapper implemented (#8) * FastText implemented * Some fixes * Fix shapes for wordsNearest * Validation of input vectors * Fixes * Fixed test * Thread tagged * Some tweaks * setContextClassLoader for DeallocatorServiceThread * Numpy format tests (#1) * Various fixes (#11) * #7852 SameDiff gather fix * #7892 SameDiff placeholder to constant conversion * #7890 validate input rank for MLN/CG init methods * Fix broken permute shape calculation * Permute and gather fixes * Tests * #7850 LogSumExp fix + test * Handful of test fixes * Empty arrays with non-scalar shapes (#10) * minor rearrangements for lambdas * empty tensors with non-scalar shapes * numpy empty tensors with non-scalar shapes * few more empty tweaks * Small fixes * conv3d signature update * micro fix in batchnorm mkldnn * Import fixes * Fix * MKL-DNN update * Small fill fix * fill with empty input + test * Fixes * Small error improvement * Fix * one special test * couple of fixes for lstm * Rewrite TFGraphMapper.getNDArrayFromTensor to be maintainable and less error prone * Fixes * FP16 * Unsigned * BFloat16 * Fill op - empty tweaks * - couple of fixes for empty arrays construction - stack updated * strided slice fix * one transform test * provide method for reducing shapeInfo in case of input array is empty * Fixed reduceAlongDimensions to use empty input properly. * couple of broadcast tests * couple of tests broadcast tests + tweak to make them pass * add check of non-empty to methods producing sub-arrays * Fixed reshapeC with zeros in shape. * complete empty check in reduce_... legacy ops * Concat and cumsum/prod * Tweak to empty shape inference on import * add empty check to the rest of reduce legacy ops * one more test * correct typo in evalReduceShapeInfoEmpty * Added tests for reduce_* ops to tests with zero shapes. * few more tests for empty reductions * Fixed strided_slice op with empty case and tests. * one more empty reduction test * Fixed strided_slice test. * add empty check to NDArray::reshapei * infOrMax * empty min/max with infinity tests * made unstack working correctly with empty arrays * few IndexReduce tests + tweaks for empty shapes * add test for empty concat * few tests fixed * Validation fix for reductions on empty shapes * Reverse fix * Reduction shape calc fixes * SameDiff.generateOutputVariable: don't use shape function to determine number of outputs * Range fix * - NDArray constructor updated for scalars/empty arrays - few tests fixed * More fixes * Empty creator fixes * concat fix * concat fix * TF import tests: allow 'both all NaN' and 'both all inf' to pass * Slice, zero fraction, and reshape fixes * transpose, gather * Zero fraction * scalar cast fix * Empty reduction axis support * few more tests fixed * Fixed input checks conforming with TF for concat op and tests. * few tests fixed * matmul scalar shape fix * Fixed checkout for data type and scalarity with concat to allow non-empty scalars with vector concats. * broadcast bool fix * few more tests * few more tests * correct evalReduceShapeInfoEmpty * argmax/argmin + tests * one more empty edge case + one more test * argmax/argmin/realdiv_bp tweaks * empty reshape test + fix * Helper fixes * Small fixes * Gather test fix * Gather test fix * Small fixes * reduce scalar zero values * scalar mean workaround * Remove debug code * along dim mean workaround * one more test * - equalsTo() tweak for empty arrays - one more test * broadcast tweaks * [WIP] Fixing outstanding issues for NLP (#9) * Avoid using not-inited objects * Test fixed. * Redundant method avoided for models like FastText * KMeans++ implementation * KMeans++ implementation * Disable parallel execution * KMeans++ * Tests * Dev branch merge (#16) * SameDiff: convertDataType and gradient check util improvements (#12) * GradCheck util improvements * StopGradient constructor + test * SameDiff: Add datatype conversion * Javadoc and add DataType.isNumerical() * Small fix * Fix SameDiff TF import test cases intermediate naming (workaround for bad default) * TFGraphTestAllHelper: check intermediates in execution order * Add missing debug listener * [WIP] lstmBlock fix + other changes (#13) - fixes lstmBlock issue - changes NDArray method reshape(), permute(), transpose() by making them return instance instead of pointer - CheckNumerics op - fixes for ReduceBool IsInfOrNan & IsFinite * Small test fix * CheckNumerics op wrapper * Fix some issues on master (#17) * Fix DataVec test issue * Fix issue with dl4j SameDiff output layer * Dtype fix for lambda layers * #7912 BertIterator dtype fix (use float32 not global default) * [WIP] Next set of CUDA stuff (#7) New CUDA implementations and improvements * bad file * Dev branch master merge (#23) * SameDiff: convertDataType and gradient check util improvements (#12) * GradCheck util improvements * StopGradient constructor + test * SameDiff: Add datatype conversion * Javadoc and add DataType.isNumerical() * Small fix * Fix SameDiff TF import test cases intermediate naming (workaround for bad default) * TFGraphTestAllHelper: check intermediates in execution order * Add missing debug listener * [WIP] lstmBlock fix + other changes (#13) - fixes lstmBlock issue - changes NDArray method reshape(), permute(), transpose() by making them return instance instead of pointer - CheckNumerics op - fixes for ReduceBool IsInfOrNan & IsFinite * Small test fix * CheckNumerics op wrapper * Compatibility of deserialization (#18) Signed-off-by: Alexander Stoyakin <alexander.stoyakin@gmail.com> * SameDiff: add activation gradient checking support for debugging (#19) * SameDiff gradient checker: first pass on activation gradient checks * Fixes + tests for activation gradient checking * Javadoc * [WIP] Some nd4j data type corrections (#20) * Adjust data type * Set correct Data type. * Size of proper data type. * fix averaged cpu load (#22) * SameDiff ops, TF import and fixes (#24) * CheckNumerics tests + fixes + misc fixes Signed-off-by: AlexDBlack <blacka101@gmail.com> * Fake quant Signed-off-by: AlexDBlack <blacka101@gmail.com> * Fixes Signed-off-by: AlexDBlack <blacka101@gmail.com> * FakeQuantWithMinMaxArgs Signed-off-by: AlexDBlack <blacka101@gmail.com> * CheckNumerics fix Signed-off-by: AlexDBlack <blacka101@gmail.com> * Fix libnd4j ALL_INTS and ALL_FLOATS declaration (uint and bfloat types) Signed-off-by: AlexDBlack <blacka101@gmail.com> * Small fix Signed-off-by: AlexDBlack <blacka101@gmail.com> * Javadoc Signed-off-by: AlexDBlack <blacka101@gmail.com> * Exception tweak Signed-off-by: AlexDBlack <blacka101@gmail.com> * fix Signed-off-by: AlexDBlack <blacka101@gmail.com> * Fix for out of scope stack allocated var use Signed-off-by: AlexDBlack <blacka101@gmail.com> * Ignores Signed-off-by: AlexDBlack <blacka101@gmail.com> * Ignore for known failing test (already logged issue) Signed-off-by: AlexDBlack <blacka101@gmail.com> * Merge upstream to fork (#25) * Add thousand-separator commas to TotalParams (#7915) * Add thousand-separator commas to TotalParams The number of parameters can be quite large, and it would help the reading of the summary printout to have the TotalParams column & values at the bottom have thousand-separator-commas in them. * Add thousand-separator commas to MultiLayerNetwork Corresponding change to MultiLayerNetwork Signed-off-by: Jxtps Jxtps <jxtps435@gmail.com> * Update contributing and issue/PR templates (#7934) Signed-off-by: AlexDBlack <blacka101@gmail.com> * Fix link to AdaDelta paper (#7942) Fix link to AdaDelta paper hosted on matthewzeiler.com Signed-off-by: Jxtps * Fixes, and ignores for known/logged failing issues (#7943) Signed-off-by: AlexDBlack <blacka101@gmail.com> * SameDiff + DL4J/SameDiff: Multiple fixes (#28) * #7919 HDF5 attribute buffer length fix Signed-off-by: AlexDBlack <blacka101@gmail.com> * #7909 Arbiter constructor exception ux improvements Signed-off-by: AlexDBlack <blacka101@gmail.com> * #7925 RNN output layer length checks Signed-off-by: AlexDBlack <blacka101@gmail.com> * #7939 Add listener for validating inputs are not incorrectly modified Signed-off-by: AlexDBlack <blacka101@gmail.com> * #7939 Integrate NonInplaceValidationListener into tests * #7844 DL4J SameDiff fixes for variable minibatch size * DL4J SameDiff fixes - ensure gradient for input placeholder is available Signed-off-by: AlexDBlack <blacka101@gmail.com> * Tweaks to ExternalErrorsFunction - use placeholders, make more robust * Another fix * More fixes * More SameDiff/DL4J fixes * Scope out scalar array creation in BaseScalarOp * Remove debug code Signed-off-by: AlexDBlack <blacka101@gmail.com> * [WIP] Final dev branch merge (#29) * SameDiff: convertDataType and gradient check util improvements (#12) * GradCheck util improvements * StopGradient constructor + test * SameDiff: Add datatype conversion * Javadoc and add DataType.isNumerical() * Small fix * Fix SameDiff TF import test cases intermediate naming (workaround for bad default) * TFGraphTestAllHelper: check intermediates in execution order * Add missing debug listener * [WIP] lstmBlock fix + other changes (#13) - fixes lstmBlock issue - changes NDArray method reshape(), permute(), transpose() by making them return instance instead of pointer - CheckNumerics op - fixes for ReduceBool IsInfOrNan & IsFinite * Small test fix * CheckNumerics op wrapper * Compatibility of deserialization (#18) Signed-off-by: Alexander Stoyakin <alexander.stoyakin@gmail.com> * SameDiff: add activation gradient checking support for debugging (#19) * SameDiff gradient checker: first pass on activation gradient checks * Fixes + tests for activation gradient checking * Javadoc * [WIP] Some nd4j data type corrections (#20) * Adjust data type * Set correct Data type. * Size of proper data type. * fix averaged cpu load (#22) * [WIP] Multiple dataset iterators (#27) * Splitting dataset into arbitrary number * Fixes * Multiple split of iterator * Test * Test * Some fixes * signature change * one more tweak Signed-off-by: raver119 <raver119@gmail.com> * one more test for sequential use of DataSetIteratorSplitter Signed-off-by: raver119 <raver119@gmail.com> * Fixes * Fixes * one more test for Alexander Signed-off-by: raver119 <raver119@gmail.com> * Some fixes * Some fixes * one more test for Alexander Signed-off-by: raver119 <raver119@gmail.com> * minor test fix Signed-off-by: raver119 <raver119@gmail.com> * Some fixes * Some fixes * couple of assertions tweaked Signed-off-by: raver119 <raver119@gmail.com> * MDS splitter test :/ Signed-off-by: raver119 <raver119@gmail.com> * Minor refactoring * Multi dataset * Some fixes * More tests * Small number of test fixes/improvements (failures on CI) (#31) Signed-off-by: AlexDBlack <blacka101@gmail.com> * [WIP] More CUDA stuff (#26) * initial commit Signed-off-by: raver119 <raver119@gmail.com> * LRN BP CUDA Signed-off-by: raver119 <raver119@gmail.com> * less memory Signed-off-by: raver119 <raver119@gmail.com> * Fixed bug with crop_and_resize op helper. * get rid of unnecessary index-calculation dunction Signed-off-by: Yurii <yurii@skymind.io> * Fixed sort with nth_element cuda-based helper. * Refactored nth_element. * Refactored nth_element op and tests. * Modified usage of dim array with sortTad routine. * Refactored main routine of helper for non_max_image_suppression op. * non_max_image_suppression op helper with cuda kernel implementation. Initial revision. * fix vol2col cuda kernel * meh Signed-off-by: raver119 <raver119@gmail.com> * topK concept Signed-off-by: raver119 <raver119@gmail.com> * unsorted topK with scanWitdh of 1 Signed-off-by: raver119 <raver119@gmail.com> * correct vol2col tests * sorted/unsorted topK Signed-off-by: raver119 <raver119@gmail.com> * implementation and fixing col2im/col2vol * Corrected usage flags with input/output with reverse op. * dup is const now Signed-off-by: raver119 <raver119@gmail.com> * percentile op Signed-off-by: raver119 <raver119@gmail.com> * group tests for mapool2d Signed-off-by: Yurii <yurii@skymind.io> * special test for george Signed-off-by: raver119 <raver119@gmail.com> * less threads for sortTad Signed-off-by: raver119 <raver119@gmail.com> * provide conv2d for cuda Signed-off-by: Yurii <yurii@skymind.io> * remove auther in sort tad kernel code Signed-off-by: Yurii <yurii@skymind.io> * provide depthwise_conv2d for cuda Signed-off-by: Yurii <yurii@skymind.io> * - max_pooling_with_argmax - null check for special use Signed-off-by: raver119 <raver119@gmail.com> * dts cuda Signed-off-by: raver119 <raver119@gmail.com> * provide sconv2d for cuda Signed-off-by: Yurii <yurii@skymind.io> * std cuda Signed-off-by: raver119 <raver119@gmail.com> * Refactored non_max_suppression op to conform TF implementation. * Improved suppression helper. * provide pooling3d for cuda Signed-off-by: Yurii <yurii@skymind.io> * minor lstm rearrangements Signed-off-by: raver119 <raver119@gmail.com> * more of minor lstm rearrangements Signed-off-by: raver119 <raver119@gmail.com> * (bi)dynamic_rnn Signed-off-by: raver119 <raver119@gmail.com> * templates init order Signed-off-by: raver119 <raver119@gmail.com> * Refactored non_max_suppression op. * Added cuda kernel for non_max_suppression. * CPU sort by key/value Signed-off-by: raver119 <raver119@gmail.com> * CPU sort TAD by key/value Signed-off-by: raver119 <raver119@gmail.com> * CPU sort TAD by key/value tests Signed-off-by: raver119 <raver119@gmail.com> * Eliminate compiler error with cuda implementation. * - repaired gradCheck in cuda - provide conv2d_bp for cuda Signed-off-by: Yurii <yurii@skymind.io> * missed signature Signed-off-by: raver119 <raver119@gmail.com> * provide depthwise_conv2d_bp for cuda Signed-off-by: Yurii <yurii@skymind.io> * Implementation of lup helper with cuda kernel. Initial commit. * further work on backprops for convolutions Signed-off-by: Yurii <yurii@skymind.io> * CUDA linear sort by key/val Signed-off-by: raver119 <raver119@gmail.com> * CUDA tad sort by key/val Signed-off-by: raver119 <raver119@gmail.com> * start providing of backprop for pooling2d/3d Signed-off-by: Yurii <yurii@skymind.io> * Added atomicAdd for bool datatype. * dynamic partition concept Signed-off-by: raver119 <raver119@gmail.com> * dynamic partition concept Signed-off-by: raver119 <raver119@gmail.com> * dynamic partition scalar CUDA Signed-off-by: raver119 <raver119@gmail.com> * important comment Signed-off-by: raver119 <raver119@gmail.com> * fix pooling2d/3d backprop helpers Signed-off-by: Yurii <yurii@skymind.io> * Added non-linear test with dynamic_partition. * Improved test for dynamic_partition. * dynamic_partition TAD concept Signed-off-by: raver119 <raver119@gmail.com> * - dynamic_partition TAD CUDA impl - dynamic_partition TAD CPU fix Signed-off-by: raver119 <raver119@gmail.com> * - rewrite cpu code for usampling2d/3d - write cuda code for usampling2d/3d Signed-off-by: Yurii <yurii@skymind.io> * dynamic_stitch CUDA vector case Signed-off-by: raver119 <raver119@gmail.com> * dynamic_stitch CUDA TAD case concept Signed-off-by: raver119 <raver119@gmail.com> * dynamic_stitch CUDA TAD case impl Signed-off-by: raver119 <raver119@gmail.com> * Added tests for dynamic_stitch 3D-4D cases. * minor tests tweaks Signed-off-by: raver119 <raver119@gmail.com> * Fixed type check for dynamic stitch. * min/max bp Signed-off-by: raver119 <raver119@gmail.com> * rewrite code for upsampling2d/3d cpu Signed-off-by: Yurii <yurii@skymind.io> * reduce min/max/norm_max bp Signed-off-by: raver119 <raver119@gmail.com> * lup implementation. Additional enhancements. * provide code for upsamling2d/3d backprop Signed-off-by: Yurii <yurii@skymind.io> * weightedCrossEntropyWithLogits Signed-off-by: raver119 <raver119@gmail.com> * Fixed template math atomicMul for 64bit ints. * Refactored dynamic_partition_bp op. * inverseBroadcast fix Signed-off-by: raver119 <raver119@gmail.com> * DynamicPartitionBP test datatype fixed. * - nd4j_atomicMul Windows fix - cpu/NDArrayLambda.hpp excluded from CUDA Signed-off-by: raver119 <raver119@gmail.com>
2019-06-27 17:37:04 +02:00
////////////////////////////////////////////////////////////////////////
void* NDArray::specialBuffer() {
if (_buffer->special() == nullptr)
return getBuffer();
// FIXME: this should be fixed once CUDA backend added
return static_cast<int8_t*>(_buffer->special()) + (_offset * sizeOfT());
}
////////////////////////////////////////////////////////////////////////
void* NDArray::getSpecialBuffer() const {
if (_buffer->special() == nullptr)
return getBuffer();
// FIXME: this should be fixed once CUDA backend added
return static_cast<int8_t*>(_buffer->special()) + (_offset * sizeOfT());
}
2019-06-06 14:21:15 +02:00
//////////////////////////////////////////////////////////////////////////
// change an array by repeating it the number of times given by reps.
NDArray NDArray::tile(const std::vector<Nd4jLong>& reps) const {
[WIP] more CUDA stuff (#57) * initial commit Signed-off-by: raver119 <raver119@gmail.com> * Added gradcheck test for dynamic_partition_bp op. * - implementation of dilation op (cpu and cuda) Signed-off-by: Yurii <yurii@skymind.io> * Fixed broadcast_dynamic_shape 1D case and tests. * Fixed usage of default integer arguments. * Fixed dynamic_partition_bp op and tests. * Eliminated test with grad check for dynamic_partition_bp op. * start working on cuda svd - porting available corresponding api from cuSOLVER library Signed-off-by: Yurii <yurii@skymind.io> * provide prelu_bp Signed-off-by: Yurii <yurii@skymind.io> * - provide gruCell_bp (old version ??) Signed-off-by: Yurii <yurii@skymind.io> * - polishing cumsum_bp and cumprod_bp tests Signed-off-by: Yurii <yurii@skymind.io> * provide sparseSoftmaxCrossEntropyWithLogits and sparseSoftmaxCrossEntropyWithLogits_grad Signed-off-by: Yurii <yurii@skymind.io> * Fixed atomicMul with float input/output * implementation of cuda kernel for triu_bp operation Signed-off-by: Yurii <yurii@skymind.io> * Refactored lup helper to add parrallel computing. * cusolver libraries Signed-off-by: raver119 <raver119@gmail.com> * uncomment cuSolver APIs in svd.cu Signed-off-by: Yurii <yurii@skymind.io> * cusolver var Signed-off-by: raver119 <raver119@gmail.com> * - further work on cuSolver svd Signed-off-by: Yurii <yurii@skymind.io> * Implement usage of cuda solver to LUP decomposition. * - correct naames in lup functions Signed-off-by: Yurii <yurii@skymind.io> * correct svdQR cuda Signed-off-by: Yurii <yurii@skymind.io> * - provide transpositions of input matrices in case of c order in svdCudaQR Signed-off-by: Yurii <yurii@skymind.io> * Fixed implementation issues with LUP usign cuda solver. * Implementation of matrix_determinant helper with cuda kernels. Working revision. * Implemented log_matrix_determinant helper with cuda kernels. * - implementation of batched cuda svd Signed-off-by: Yurii <yurii@skymind.io> * Refactored cholesky helper and implementation of cuda solver cholesky batch. * - implementation of cuda kernel for tile bp Signed-off-by: Yurii <yurii@skymind.io> * Implementation of cholesky and logdet with cuda kernels. * - implementation of cuda kernel for sru_bidirectional Signed-off-by: Yurii <yurii@skymind.io> * Fixed cholesky helper. * Cholesky op helper implementation. Working double-based cublas implementation. * bad import excluded Signed-off-by: raver119 <raver119@gmail.com> * Finished with cuda implementation of cholesky helper and tests. * - implementation of cuda kernel for sru_bidirectional_backprop operation Signed-off-by: Yurii <yurii@skymind.io> * Implementation of matrix_inverse op helper with cuda kernels. The first revision. * - start working on gruCell_bp Signed-off-by: Yurii <yurii@skymind.io> * Implementation of matrix_inverse helper. * - further work on new gruCell_bp Signed-off-by: Yurii <yurii@skymind.io> * cuBLAS related fixes Signed-off-by: raver119 <raver119@gmail.com> * calculateOutputShapes() now passes device buffers as well Signed-off-by: raver119 <raver119@gmail.com> * special concat/average/accumulate init host pointers now Signed-off-by: raver119 <raver119@gmail.com> * few more tweaks Signed-off-by: raver119 <raver119@gmail.com> * additional CudaDataBufferFactory signatures certain for data types Signed-off-by: raver119 <raver119@gmail.com> * cuSolver host buffer Signed-off-by: raver119 <raver119@gmail.com> * buffer to buffer memcpy host ptr allocation Signed-off-by: raver119 <raver119@gmail.com>
2019-07-12 10:51:51 +02:00
const int repsSize = reps.size();
2019-06-06 14:21:15 +02:00
int product = 1;
for(const auto& item : reps)
product *= item;
if(product == 0)
throw std::runtime_error("NDArray::tile method: one of the elements in reps array is zero !");
int rankOld = rankOf();
[WIP] more CUDA stuff (#57) * initial commit Signed-off-by: raver119 <raver119@gmail.com> * Added gradcheck test for dynamic_partition_bp op. * - implementation of dilation op (cpu and cuda) Signed-off-by: Yurii <yurii@skymind.io> * Fixed broadcast_dynamic_shape 1D case and tests. * Fixed usage of default integer arguments. * Fixed dynamic_partition_bp op and tests. * Eliminated test with grad check for dynamic_partition_bp op. * start working on cuda svd - porting available corresponding api from cuSOLVER library Signed-off-by: Yurii <yurii@skymind.io> * provide prelu_bp Signed-off-by: Yurii <yurii@skymind.io> * - provide gruCell_bp (old version ??) Signed-off-by: Yurii <yurii@skymind.io> * - polishing cumsum_bp and cumprod_bp tests Signed-off-by: Yurii <yurii@skymind.io> * provide sparseSoftmaxCrossEntropyWithLogits and sparseSoftmaxCrossEntropyWithLogits_grad Signed-off-by: Yurii <yurii@skymind.io> * Fixed atomicMul with float input/output * implementation of cuda kernel for triu_bp operation Signed-off-by: Yurii <yurii@skymind.io> * Refactored lup helper to add parrallel computing. * cusolver libraries Signed-off-by: raver119 <raver119@gmail.com> * uncomment cuSolver APIs in svd.cu Signed-off-by: Yurii <yurii@skymind.io> * cusolver var Signed-off-by: raver119 <raver119@gmail.com> * - further work on cuSolver svd Signed-off-by: Yurii <yurii@skymind.io> * Implement usage of cuda solver to LUP decomposition. * - correct naames in lup functions Signed-off-by: Yurii <yurii@skymind.io> * correct svdQR cuda Signed-off-by: Yurii <yurii@skymind.io> * - provide transpositions of input matrices in case of c order in svdCudaQR Signed-off-by: Yurii <yurii@skymind.io> * Fixed implementation issues with LUP usign cuda solver. * Implementation of matrix_determinant helper with cuda kernels. Working revision. * Implemented log_matrix_determinant helper with cuda kernels. * - implementation of batched cuda svd Signed-off-by: Yurii <yurii@skymind.io> * Refactored cholesky helper and implementation of cuda solver cholesky batch. * - implementation of cuda kernel for tile bp Signed-off-by: Yurii <yurii@skymind.io> * Implementation of cholesky and logdet with cuda kernels. * - implementation of cuda kernel for sru_bidirectional Signed-off-by: Yurii <yurii@skymind.io> * Fixed cholesky helper. * Cholesky op helper implementation. Working double-based cublas implementation. * bad import excluded Signed-off-by: raver119 <raver119@gmail.com> * Finished with cuda implementation of cholesky helper and tests. * - implementation of cuda kernel for sru_bidirectional_backprop operation Signed-off-by: Yurii <yurii@skymind.io> * Implementation of matrix_inverse op helper with cuda kernels. The first revision. * - start working on gruCell_bp Signed-off-by: Yurii <yurii@skymind.io> * Implementation of matrix_inverse helper. * - further work on new gruCell_bp Signed-off-by: Yurii <yurii@skymind.io> * cuBLAS related fixes Signed-off-by: raver119 <raver119@gmail.com> * calculateOutputShapes() now passes device buffers as well Signed-off-by: raver119 <raver119@gmail.com> * special concat/average/accumulate init host pointers now Signed-off-by: raver119 <raver119@gmail.com> * few more tweaks Signed-off-by: raver119 <raver119@gmail.com> * additional CudaDataBufferFactory signatures certain for data types Signed-off-by: raver119 <raver119@gmail.com> * cuSolver host buffer Signed-off-by: raver119 <raver119@gmail.com> * buffer to buffer memcpy host ptr allocation Signed-off-by: raver119 <raver119@gmail.com>
2019-07-12 10:51:51 +02:00
int diff = rankOld - repsSize;
2019-06-06 14:21:15 +02:00
if(product==1) { // in this case 2 possibilities are present: just reshape or nothing to do
NDArray result(*this);
if(diff < 0) { // reshape to higher dimension
[WIP] more CUDA stuff (#57) * initial commit Signed-off-by: raver119 <raver119@gmail.com> * Added gradcheck test for dynamic_partition_bp op. * - implementation of dilation op (cpu and cuda) Signed-off-by: Yurii <yurii@skymind.io> * Fixed broadcast_dynamic_shape 1D case and tests. * Fixed usage of default integer arguments. * Fixed dynamic_partition_bp op and tests. * Eliminated test with grad check for dynamic_partition_bp op. * start working on cuda svd - porting available corresponding api from cuSOLVER library Signed-off-by: Yurii <yurii@skymind.io> * provide prelu_bp Signed-off-by: Yurii <yurii@skymind.io> * - provide gruCell_bp (old version ??) Signed-off-by: Yurii <yurii@skymind.io> * - polishing cumsum_bp and cumprod_bp tests Signed-off-by: Yurii <yurii@skymind.io> * provide sparseSoftmaxCrossEntropyWithLogits and sparseSoftmaxCrossEntropyWithLogits_grad Signed-off-by: Yurii <yurii@skymind.io> * Fixed atomicMul with float input/output * implementation of cuda kernel for triu_bp operation Signed-off-by: Yurii <yurii@skymind.io> * Refactored lup helper to add parrallel computing. * cusolver libraries Signed-off-by: raver119 <raver119@gmail.com> * uncomment cuSolver APIs in svd.cu Signed-off-by: Yurii <yurii@skymind.io> * cusolver var Signed-off-by: raver119 <raver119@gmail.com> * - further work on cuSolver svd Signed-off-by: Yurii <yurii@skymind.io> * Implement usage of cuda solver to LUP decomposition. * - correct naames in lup functions Signed-off-by: Yurii <yurii@skymind.io> * correct svdQR cuda Signed-off-by: Yurii <yurii@skymind.io> * - provide transpositions of input matrices in case of c order in svdCudaQR Signed-off-by: Yurii <yurii@skymind.io> * Fixed implementation issues with LUP usign cuda solver. * Implementation of matrix_determinant helper with cuda kernels. Working revision. * Implemented log_matrix_determinant helper with cuda kernels. * - implementation of batched cuda svd Signed-off-by: Yurii <yurii@skymind.io> * Refactored cholesky helper and implementation of cuda solver cholesky batch. * - implementation of cuda kernel for tile bp Signed-off-by: Yurii <yurii@skymind.io> * Implementation of cholesky and logdet with cuda kernels. * - implementation of cuda kernel for sru_bidirectional Signed-off-by: Yurii <yurii@skymind.io> * Fixed cholesky helper. * Cholesky op helper implementation. Working double-based cublas implementation. * bad import excluded Signed-off-by: raver119 <raver119@gmail.com> * Finished with cuda implementation of cholesky helper and tests. * - implementation of cuda kernel for sru_bidirectional_backprop operation Signed-off-by: Yurii <yurii@skymind.io> * Implementation of matrix_inverse op helper with cuda kernels. The first revision. * - start working on gruCell_bp Signed-off-by: Yurii <yurii@skymind.io> * Implementation of matrix_inverse helper. * - further work on new gruCell_bp Signed-off-by: Yurii <yurii@skymind.io> * cuBLAS related fixes Signed-off-by: raver119 <raver119@gmail.com> * calculateOutputShapes() now passes device buffers as well Signed-off-by: raver119 <raver119@gmail.com> * special concat/average/accumulate init host pointers now Signed-off-by: raver119 <raver119@gmail.com> * few more tweaks Signed-off-by: raver119 <raver119@gmail.com> * additional CudaDataBufferFactory signatures certain for data types Signed-off-by: raver119 <raver119@gmail.com> * cuSolver host buffer Signed-off-by: raver119 <raver119@gmail.com> * buffer to buffer memcpy host ptr allocation Signed-off-by: raver119 <raver119@gmail.com>
2019-07-12 10:51:51 +02:00
std::vector<Nd4jLong> shapeNew = reps; // there is requirement to have unities at first "diff" positions of new shape
2019-06-06 14:21:15 +02:00
memcpy(&shapeNew[-diff], result.getShapeInfo()+1, rankOld * sizeof(Nd4jLong)); // put old shape numbers at rest of positions
result.reshapei(ordering(), shapeNew);
}
return result; // nothing to do, if diff >= 0 -> identity tile
}
// evaluate shapeInfo for resulting array
auto newShapeInfo = ShapeUtils::evalTileShapeInfo(*this, reps, getContext()->getWorkspace());
// create new buffer, in any case the memory amount new buffer points to is bigger then those for old _buffer
std::shared_ptr<DataBuffer> newBuff = std::make_shared<DataBuffer>(shape::length(newShapeInfo) * sizeOfT(), dataType(), getContext()->getWorkspace());
// assign new shape and new buffer to resulting array
NDArray result(newBuff, ShapeDescriptor(newShapeInfo), getContext());
// fill newBuff, loop through all elements of newBuff
// looping through _buffer goes automatically by means of getSubArrayIndex applying
const auto resultLen = result.lengthOf();
auto xType = this->dataType();
if(result.ordering() == 'c') { // ews == 1 always here
PRAGMA_OMP_PARALLEL_FOR_SIMD
for(Nd4jLong i = 0; i < resultLen; ++i) {
auto yOffset = shape::subArrayOffset(i, newShapeInfo, getShapeInfo());
BUILD_SINGLE_SELECTOR(xType, this->template templatedAssign, (result.getBuffer(), i, this->getBuffer(), yOffset), LIBND4J_TYPES);
}
}
else {
PRAGMA_OMP_PARALLEL_FOR_SIMD
for(Nd4jLong i=0; i<resultLen; ++i) {
2019-06-06 14:21:15 +02:00
auto xOffset = result.getOffset(i);
auto yOffset = shape::subArrayOffset(i, newShapeInfo, getShapeInfo());
BUILD_SINGLE_SELECTOR(xType, this->template templatedAssign, (result.getBuffer(), xOffset, this->getBuffer(), yOffset), LIBND4J_TYPES);
}
}
result.tickWriteHost();
return result;
}
//////////////////////////////////////////////////////////////////////////
// change an array by repeating it the number of times given by reps.
void NDArray::tile(const std::vector<Nd4jLong>& reps, NDArray& target) const {
// evaluate true tile shapeInfo for comparison with target shapeInfo
auto newShapeInfo = ShapeUtils::evalTileShapeInfo(*this, reps, getContext()->getWorkspace());
if(!shape::equalsSoft(newShapeInfo, target.getShapeInfo())) {
delete []newShapeInfo;
throw std::runtime_error("NDArray::tile method - shapeInfo of target array is not suitable for tile operation !");
}
// fill newBuff, loop through all elements of newBuff
// looping through _buffer goes automatically by means of getSubArrayIndex applying
const int ews = target.ews();
const int targetLen = target.lengthOf();
if(target.ordering() == 'c' && ews == 1) { // ews == 1 always here
//#pragma omp parallel for simd if(targetLen > Environment::getInstance()->elementwiseThreshold()) schedule(guided)
for(Nd4jLong i=0; i<targetLen; ++i) {
auto yOffset = shape::subArrayOffset(i, target.getShapeInfo(), getShapeInfo());
BUILD_DOUBLE_SELECTOR(target.dataType(), dataType(), templatedDoubleAssign, (target.getBuffer(), i, getBuffer(), yOffset), LIBND4J_TYPES, LIBND4J_TYPES);
}
}
else if(target.ordering() == 'c' && ews > 1) {
for(Nd4jLong i=0; i<targetLen; ++i) {
2019-06-06 14:21:15 +02:00
auto yOffset = shape::subArrayOffset(i, target.getShapeInfo(), getShapeInfo());
BUILD_DOUBLE_SELECTOR(target.dataType(), dataType(), templatedDoubleAssign, (target.getBuffer(), i*ews, getBuffer(), yOffset), LIBND4J_TYPES, LIBND4J_TYPES);
}
}
else {
for(Nd4jLong i=0; i<targetLen; ++i) {
2019-06-06 14:21:15 +02:00
auto xOffset = target.getOffset(i);
auto yOffset = shape::subArrayOffset(i, target.getShapeInfo(), getShapeInfo());
BUILD_DOUBLE_SELECTOR(target.dataType(), dataType(), templatedDoubleAssign, (target.getBuffer(), xOffset, getBuffer(), yOffset), LIBND4J_TYPES, LIBND4J_TYPES);
}
}
}
//////////////////////////////////////////////////////////////////////////
void NDArray::tile(NDArray& target) const {
if(rankOf() > target.rankOf())
throw std::runtime_error("NDArray::tile method - rank of target array must be bigger or equal to the rank of this array !");
if(!ShapeUtils::areShapesBroadcastable(*this, target))
throw std::runtime_error("NDArray::tile method - shapeInfo of target array is not suitable for tile operation !");
// fill newBuff, loop through all elements of newBuff
// looping through _buffer goes automatically by means of getSubArrayIndex applying
const auto ews = target.ews();
const auto targetLen = target.lengthOf();
if(target.ordering() == 'c' && ews == 1) { // ews == 1 always here
for (Nd4jLong i = 0; i < targetLen; ++i) {
2019-06-06 14:21:15 +02:00
auto yOffset = shape::subArrayOffset(i, target.getShapeInfo(), getShapeInfo());
BUILD_DOUBLE_SELECTOR(target.dataType(), dataType(), templatedDoubleAssign, (target.getBuffer(), i, getBuffer(), yOffset), LIBND4J_TYPES, LIBND4J_TYPES);
}
}
else if(target.ordering() == 'c' && ews > 1) {
for(Nd4jLong i=0; i<targetLen; ++i) {
2019-06-06 14:21:15 +02:00
auto yOffset = shape::subArrayOffset(i, target.getShapeInfo(), getShapeInfo());
BUILD_DOUBLE_SELECTOR(target.dataType(), dataType(), templatedDoubleAssign, (target.getBuffer(), i*ews, getBuffer(), yOffset), LIBND4J_TYPES, LIBND4J_TYPES);
}
}
else {
for(Nd4jLong i=0; i<targetLen; ++i) {
2019-06-06 14:21:15 +02:00
auto xOffset = target.getOffset(i);
auto yOffset = shape::subArrayOffset(i, target.getShapeInfo(), getShapeInfo());
BUILD_DOUBLE_SELECTOR(target.dataType(), dataType(), templatedDoubleAssign, (target.getBuffer(), xOffset, getBuffer(), yOffset), LIBND4J_TYPES, LIBND4J_TYPES);
}
}
}
//////////////////////////////////////////////////////////////////////////
// create new array by repeating it the number of times given by reps
NDArray* NDArray::repeat(int dimension, const std::vector<Nd4jLong>& repeats) const {
auto outShape = ShapeUtils::evalRepeatShape(dimension, repeats, *this);
// the size of outShape == rank
int rank = rankOf(); // = outShape.size()
auto ret = new NDArray('c', outShape, dataType(), getContext());
auto retArrs = ret->allTensorsAlongDimension({dimension});
auto thisArrs = this->allTensorsAlongDimension({dimension});
auto repeatDelta = shape::prodLong(outShape.data(), rank) / this->lengthOf();
auto numTads = retArrs->size();
for (int i = 0; i < numTads; i++) {
auto thisTensor = thisArrs->at(i);
auto retTensor = retArrs->at(i);
Nd4jLong retIdx = 0;
for (Nd4jLong k = 0; k < thisTensor->lengthOf(); k++) {
auto s = thisTensor->e(k);
for (Nd4jLong j = 0; j < repeatDelta; j++)
retTensor->p(retIdx++, s);
}
}
delete retArrs;
delete thisArrs;
return ret;
}
//////////////////////////////////////////////////////////////////////////
// fill array by repeating it the number of times given by reps
void NDArray::repeat(int dimension, NDArray& target) const {
if(dimension < 0)
dimension += rankOf();
if(rankOf() != target.rankOf())
throw std::invalid_argument("NDArray::repeat(int dimension, NDArray& target) method: wrong rank of target array it must be equal to this array rank!");
Nd4jLong repeatDelta = target.sizeAt(dimension) / sizeAt(dimension);
if(repeatDelta == 0)
throw std::invalid_argument("NDArray::repeat(int dimension, NDArray& target) method: wrong shape of target array!");
std::vector<int> dimsToExclude = ShapeUtils::evalDimsToExclude(rankOf(), {dimension});
const Nd4jLong numTads = ShapeUtils::getNumOfSubArrs(getShapeInfo(), dimsToExclude);
for (int i = 0; i < numTads; i++) {
auto thisTensor = (*this)(i, dimsToExclude);
auto retTensor = target(i, dimsToExclude);
int tensorLength = thisTensor.lengthOf();
int retIdx = 0;
if (isR()) {
for (int k = 0; k < tensorLength; k++) {
auto s = thisTensor.e<double>(k);
for (int j = 0; j < repeatDelta; j++) {
retTensor.p<double>(retIdx++, s);
}
}
} else {
for (int k = 0; k < tensorLength; k++) {
auto s = thisTensor.e<Nd4jLong>(k);
for (int j = 0; j < repeatDelta; j++) {
retTensor.p<Nd4jLong>(retIdx++, s);
}
}
}
}
}
//////////////////////////////////////////////////////////////////////////
#ifndef __JAVACPP_HACK__
Dev branch merge: dev_20190606 (#7904) * correct logsoftmax looss (#2) * Small SameDiff listener fix (#4) * Various fixes (#6) * #7839 Fix for asXMatrix and tests * #7866 EmbeddingSequenceLayer dtype fix + test * #7856 SameDiff save/load stream methods * #7859 RegressionEvaluation rank 4 fix + tests + axis configuration * EvaluationBinary 3d/4d * More evaluation 3d/4d tests * #7847 Evaluation empty checks * Small test ifx * #7848 Fix median edge case * Improve DL4J samediff layer tests * [WIP] FastText wrapper implemented (#8) * FastText implemented * Some fixes * Fix shapes for wordsNearest * Validation of input vectors * Fixes * Fixed test * Thread tagged * Some tweaks * setContextClassLoader for DeallocatorServiceThread * Numpy format tests (#1) * Various fixes (#11) * #7852 SameDiff gather fix * #7892 SameDiff placeholder to constant conversion * #7890 validate input rank for MLN/CG init methods * Fix broken permute shape calculation * Permute and gather fixes * Tests * #7850 LogSumExp fix + test * Handful of test fixes * Empty arrays with non-scalar shapes (#10) * minor rearrangements for lambdas * empty tensors with non-scalar shapes * numpy empty tensors with non-scalar shapes * few more empty tweaks * Small fixes * conv3d signature update * micro fix in batchnorm mkldnn * Import fixes * Fix * MKL-DNN update * Small fill fix * fill with empty input + test * Fixes * Small error improvement * Fix * one special test * couple of fixes for lstm * Rewrite TFGraphMapper.getNDArrayFromTensor to be maintainable and less error prone * Fixes * FP16 * Unsigned * BFloat16 * Fill op - empty tweaks * - couple of fixes for empty arrays construction - stack updated * strided slice fix * one transform test * provide method for reducing shapeInfo in case of input array is empty * Fixed reduceAlongDimensions to use empty input properly. * couple of broadcast tests * couple of tests broadcast tests + tweak to make them pass * add check of non-empty to methods producing sub-arrays * Fixed reshapeC with zeros in shape. * complete empty check in reduce_... legacy ops * Concat and cumsum/prod * Tweak to empty shape inference on import * add empty check to the rest of reduce legacy ops * one more test * correct typo in evalReduceShapeInfoEmpty * Added tests for reduce_* ops to tests with zero shapes. * few more tests for empty reductions * Fixed strided_slice op with empty case and tests. * one more empty reduction test * Fixed strided_slice test. * add empty check to NDArray::reshapei * infOrMax * empty min/max with infinity tests * made unstack working correctly with empty arrays * few IndexReduce tests + tweaks for empty shapes * add test for empty concat * few tests fixed * Validation fix for reductions on empty shapes * Reverse fix * Reduction shape calc fixes * SameDiff.generateOutputVariable: don't use shape function to determine number of outputs * Range fix * - NDArray constructor updated for scalars/empty arrays - few tests fixed * More fixes * Empty creator fixes * concat fix * concat fix * TF import tests: allow 'both all NaN' and 'both all inf' to pass * Slice, zero fraction, and reshape fixes * transpose, gather * Zero fraction * scalar cast fix * Empty reduction axis support * few more tests fixed * Fixed input checks conforming with TF for concat op and tests. * few tests fixed * matmul scalar shape fix * Fixed checkout for data type and scalarity with concat to allow non-empty scalars with vector concats. * broadcast bool fix * few more tests * few more tests * correct evalReduceShapeInfoEmpty * argmax/argmin + tests * one more empty edge case + one more test * argmax/argmin/realdiv_bp tweaks * empty reshape test + fix * Helper fixes * Small fixes * Gather test fix * Gather test fix * Small fixes * reduce scalar zero values * scalar mean workaround * Remove debug code * along dim mean workaround * one more test * - equalsTo() tweak for empty arrays - one more test * broadcast tweaks
2019-06-15 13:34:34 +02:00
#include "NDArrayLambda.hpp"
2019-06-06 14:21:15 +02:00
#endif
/*
#ifndef __CLION_IDE__
#include "NDArray.macro"
#endif
*/
}
#endif