Commit Graph

8 Commits (18d4eaa68deb9819525b092fe6091e66ecfbbb6b)

Author SHA1 Message Date
Yurii Shyrma 78934c17ad
profiling of stack and unstack ops (#261)
* - profiling of stack and unstack ops

Signed-off-by: Yurii <iuriish@yahoo.com>

* - fix bug in cpu concat op

Signed-off-by: Yurii <iuriish@yahoo.com>

* - correction of cuda stack and unstack

Signed-off-by: Yurii <iuriish@yahoo.com>

* - change shape.h method which operates with unity dimensions strides

Signed-off-by: Yurii <iuriish@yahoo.com>

* - rearrange stack tests

Signed-off-by: Yurii <iuriish@yahoo.com>

* - correct evaluation of smallest stride for moving through contiguous axis

Signed-off-by: Yurii <iuriish@yahoo.com>

* - forgot to update signature of function strideOverContigAxis in cuda concat and split ops

Signed-off-by: Yurii <iuriish@yahoo.com>

* - remove ShapeUtils::shapeAsString method applied before input arrays validations

Signed-off-by: Yurii <iuriish@yahoo.com>

* -  further removing of ShapeUtils::shapeAsString

Signed-off-by: Yurii <iuriish@yahoo.com>

* - take sub-array shapeIndo/offset calculation out of NDArray class
- add possibility of contiguous memory copy in execTransformAny op if opNum == assign

Signed-off-by: Yurii <iuriish@yahoo.com>

* - correct test_empty_scatter_2 in EmptyTests.cpp

Signed-off-by: Yurii <iuriish@yahoo.com>

* - profiling of slice op

Signed-off-by: Yurii <iuriish@yahoo.com>

* - get rid of contiguous memcpy for some cases in concat and split ops

Signed-off-by: Yurii <iuriish@yahoo.com>

* - forgot to declare oid nd4j::SpecialMethods<T>::splitCpuGeneric

Signed-off-by: Yurii <iuriish@yahoo.com>

* - correct typo in calculation of threads in cuda split op

Signed-off-by: Yurii <iuriish@yahoo.com>

* - forgot to correct another set of threads variables in split cuda ops

Signed-off-by: Yurii <iuriish@yahoo.com>

* - further conflicts resolving

Signed-off-by: Yurii <iuriish@yahoo.com>

Co-authored-by: raver119 <raver119@gmail.com>
2020-03-03 07:32:37 +03:00
raver119 63fa3c2ef3
libnd4j polishing (#273)
* initial set of include changes

Signed-off-by: raver119 <raver119@gmail.com>

* one more tweak

Signed-off-by: raver119 <raver119@gmail.com>

* few more rearrangements

Signed-off-by: raver119 <raver119@gmail.com>

* few more rearrangements

Signed-off-by: raver119 <raver119@gmail.com>

* few more rearrangements

Signed-off-by: raver119 <raver119@gmail.com>

* cuda includes rearrangements

Signed-off-by: raver119 <raver119@gmail.com>

* java update

Signed-off-by: raver119 <raver119@gmail.com>

* = namespace changed to sd
- few CMake variables renamed with SD_ prefix

Signed-off-by: raver119 <raver119@gmail.com>

* java update

Signed-off-by: raver119 <raver119@gmail.com>

* LoopKind minor fix

Signed-off-by: raver119 <raver119@gmail.com>

* few more changes

Signed-off-by: raver119 <raver119@gmail.com>

* few more changes

Signed-off-by: raver119 <raver119@gmail.com>

* few more changes

Signed-off-by: raver119 <raver119@gmail.com>

* sanitizer is optional now

Signed-off-by: raver119 <raver119@gmail.com>

* dev tests updated

Signed-off-by: raver119 <raver119@gmail.com>

* few more changes

Signed-off-by: raver119 <raver119@gmail.com>

* last update

Signed-off-by: raver119 <raver119@gmail.com>

* java update

Signed-off-by: raver119 <raver119@gmail.com>
2020-03-02 12:49:41 +03:00
shugeo 1bb3ae4b03
Shugeo unordered map (#256)
* Refactored usage of std::map to std::unordered_map instead.

Signed-off-by: shugeo <sgazeos@gmail.com>

* Eliminated crash with wrong ShapeDescriptor hash.

Signed-off-by: shugeo <sgazeos@gmail.com>

* Eliminated crash with TadDescriptor hash.

Signed-off-by: shugeo <sgazeos@gmail.com>

* Refactored Stash hash.

Signed-off-by: shugeo <sgazeos@gmail.com>

* Refactored hashes.

Signed-off-by: shugeo <sgazeos@gmail.com>

* Refactored TadDescriptor hash and top_k mapping.

* Refactored hashes for ShapeDescriptor and TadDescriptor classes.

Signed-off-by: shugeo <sgazeos@gmail.com>

* Refactored hash for ConstantDescriptor and ShapeDescriptor classes.

Signed-off-by: shugeo <sgazeos@gmail.com>

* Fixed map using with cuda platform.

Signed-off-by: shugeo <sgazeos@gmail.com>

* - few rearrangements for hash functions
- shared openblas allowed

Signed-off-by: raver119 <raver119@gmail.com>

* exports

Signed-off-by: raver119 <raver119@gmail.com>

* exports

Signed-off-by: raver119 <raver119@gmail.com>

* Stash reverted to std::map

Signed-off-by: raver119@gmail.com <raver119@gmail.com>

* Added additional test.

Signed-off-by: shugeo <sgazeos@gmail.com>

* different maps for different compilers

Signed-off-by: raver119 <raver119@gmail.com>

* missing include

Signed-off-by: raver119 <raver119@gmail.com>

* fix the leak

Signed-off-by: raver119 <raver119@gmail.com>

Co-authored-by: raver119 <raver119@gmail.com>
2020-02-24 07:51:01 +03:00
Yurii Shyrma fe47f52896
Oleh tenzor mmul (#231)
* Libnd4j: TensorMMul backprop op #8174, raw implementation

Signed-off-by: Oleg <oleg.semeniv@gmail.com>

* Libnd4j: TensorMMul backprop op #8174 merge master and some corrections

Signed-off-by: Oleg <oleg.semeniv@gmail.com>

* Libnd4j: TensorMMul backprop op #8174 algorithm update, need testing, sync with  master

* Libnd4j: TensorMMul backprop op #8174 fixed incorrect B axes calculation

Signed-off-by: Oleg <oleg.semeniv@gmail.com>

* Libnd4j: TensorMMul backprop op #8174 optimize axes identification and fix bug of indeces overlapping, added first test. need testing with different shapes

Signed-off-by: Oleg <oleg.semeniv@gmail.com>

* Libnd4j: TensorMMul backprop op #8174 some fixes and improvements need more testing

Signed-off-by: Oleg <oleg.semeniv@gmail.com>

* Libnd4j: TensorMMul backprop op #8174 fixed order of matrix multiply

Signed-off-by: Oleg <oleg.semeniv@gmail.com>

* Libnd4j: TensorMMul backprop op #8174 fixed issue of incorrect axes definition, add tests based on TF, need additional testing for case dLdC not equal 1

Signed-off-by: Oleg <oleg.semeniv@gmail.com>

* Libnd4j: TensorMMul backprop op #8174 fixed scalar case add test

Signed-off-by: Oleg <oleg.semeniv@gmail.com>

* Libnd4j: TensorMMul backprop op #8174 fixed bp algorithm, axes definition, need some mode testing with different orders combination f,c; c,f f,f and add some checks for inputs

Signed-off-by: Oleg <oleg.semeniv@gmail.com>

* Libnd4j: TensorMMul backprop op #8174 some checks and corrections added tests, exists the problem with different input orders support A-f B-c and A-f B-f

Signed-off-by: Oleg <oleg.semeniv@gmail.com>

* Libnd4j: TensorMMul backprop op #8174 sync master

Signed-off-by: Oleg <oleg.semeniv@gmail.com>

* - correct bug in MmulHelper::tensorDot(a, b, c, axes_a, axes_b,permutForC)

Signed-off-by: Yurii <iuriish@yahoo.com>

* Libnd4j: TensorMMul backprop op #8174 code clean up and refactoring

Signed-off-by: Oleg <oleg.semeniv@gmail.com>

* - add check for linspase ordered permutations in ShapeUtils::evalShapeForTensorDot

Signed-off-by: Yurii <iuriish@yahoo.com>

* - provide additional code in shape::reshape stuff in order to reduce amount of allocation/copy operations during reshaping procedure

Signed-off-by: Yurii <iuriish@yahoo.com>

* - further work on problem of wrong shape evaluation during permute/reshape procedures

Signed-off-by: Yurii <iuriish@yahoo.com>

* - still looking for bug reason in reshape/permute stuff

Signed-off-by: Yurii <iuriish@yahoo.com>

* - correct bug in transform cuda native ops

Signed-off-by: Yurii <iuriish@yahoo.com>

* - correct bug in NDArray::assign

Signed-off-by: Yurii <iuriish@yahoo.com>

* - remove old shape::reshape stuff

Signed-off-by: Yurii <iuriish@yahoo.com>

* - add possibility to disable copy of old buffer to new buffer during reshape operation in NDArray class

Signed-off-by: Yurii <iuriish@yahoo.com>

* - correct bug in tensorDot which had to do with wrong pointers assigments

Signed-off-by: Yurii <iuriish@yahoo.com>

Co-authored-by: Oleh <oleg.semeniv@gmail.com>
2020-02-13 20:33:54 +03:00
raver119 7abc574eeb
Snapshot update (#8194)
* fix double consumption of rng on cpu

Signed-off-by: raver119 <raver119@gmail.com>

* Shyrma docs (#222)

* - documenting and profiling matrix_set_diag cuda kernel

Signed-off-by: Yurii <yurii@skymind.io>

* - correct formula of pnorm pooling in cuda 2d/3d kernels
- remove helper matrix_diag which duplicates work of helper matrix_set_diag

Signed-off-by: Yurii <yurii@skymind.io>

* cublasHandle sharing + lock

Signed-off-by: raver119 <raver119@gmail.com>

* cublasHandle sharing + lock

Signed-off-by: raver119 <raver119@gmail.com>

* Documentation from serialization/deserialization in NLP (#221)

* refactoring

Signed-off-by: Alexander Stoyakin <alexander.stoyakin@gmail.com>

* Javadocs

Signed-off-by: Alexander Stoyakin <alexander.stoyakin@gmail.com>

* Javadoc fixed

Signed-off-by: Alexander Stoyakin <alexander.stoyakin@gmail.com>

* Cleanup

Signed-off-by: Alexander Stoyakin <alexander.stoyakin@gmail.com>

* dedicated lock for getCudaCublasHandle

Signed-off-by: raver119 <raver119@gmail.com>

* Small fixes (#223)

Signed-off-by: AlexDBlack <blacka101@gmail.com>

* ELU DL4J fixes (#224)

Signed-off-by: AlexDBlack <blacka101@gmail.com>

* javadoc (#225)

Signed-off-by: Robert Altena <Rob@Ra-ai.com>

* Small test compilation fix (#226)

Signed-off-by: AlexDBlack <blacka101@gmail.com>

* #8182 remove spark version suffix (#227)

Signed-off-by: AlexDBlack <blacka101@gmail.com>

* [WIP] Thread safety (#229)

* sync after cublas*gemm

Signed-off-by: raver119 <raver119@gmail.com>

* mutex for CublasHelper

Signed-off-by: raver119 <raver119@gmail.com>

* don't store cublasHandle in LaunchContext, it's per-device anyway

Signed-off-by: raver119 <raver119@gmail.com>

* some printout

Signed-off-by: raver119 <raver119@gmail.com>

* check for field instead

Signed-off-by: raver119 <raver119@gmail.com>

* pew-pew

Signed-off-by: raver119 <raver119@gmail.com>

* don't release ContextBuffers until device changed

Signed-off-by: raver119 <raver119@gmail.com>

* small tweak

Signed-off-by: raver119 <raver119@gmail.com>

* some logging in sgemm

Signed-off-by: raver119 <raver119@gmail.com>

* stream sync

Signed-off-by: raver119 <raver119@gmail.com>

* some more logging

Signed-off-by: raver119 <raver119@gmail.com>

* some more error checks

Signed-off-by: raver119 <raver119@gmail.com>

* one fancy test

Signed-off-by: raver119 <raver119@gmail.com>

* one fancy test

Signed-off-by: raver119 <raver119@gmail.com>

* minor AffinityManager fix

Signed-off-by: raver119 <raver119@gmail.com>

* cudaEvent error logging improvement

Signed-off-by: raver119 <raver119@gmail.com>

* ConstantHelper thread safety

Signed-off-by: raver119 <raver119@gmail.com>

* - minor corrections in ConstantTadHelper

Signed-off-by: Yurii <yurii@skymind.io>

* ConstantShapeHelper thread safety

Signed-off-by: raver119 <raver119@gmail.com>

* ConstantTadHelper.cu updated

Signed-off-by: raver119 <raver119@gmail.com>

* logging off

Signed-off-by: raver119 <raver119@gmail.com>

* logging off

Signed-off-by: raver119 <raver119@gmail.com>
2019-09-03 22:02:02 +03:00
raver119 763a225c6a [WIP] More of CUDA operations (#69)
* initial commit

Signed-off-by: raver119 <raver119@gmail.com>

* - gruCell_bp further

Signed-off-by: Yurii <yurii@skymind.io>

* - further work on gruCell_bp

Signed-off-by: Yurii <yurii@skymind.io>

* Inverse matrix cublas implementation. Partial working revision.

* Separation of segment ops helpers. Max separation.

* Separated segment_min ops.

* Separation of segment_mean/sum/prod/sqrtN ops heleprs.

* Fixed diagonal processing with LUP decomposition.

* Modified inversion approach using current state of LU decomposition.

* Implementation of matrix_inverse op with cuda kernels. Working revision.

* Implemented sequence_mask cuda helper. Eliminated waste printf with matrix_inverse implementation. Added proper tests.

* - further work on gruCell_bp (ff/cuda)

Signed-off-by: Yurii <yurii@skymind.io>

* comment one test for gruCell_bp

Signed-off-by: Yurii <yurii@skymind.io>

* - provide cuda static_rnn

Signed-off-by: Yurii <yurii@skymind.io>

* Refactored random_shuffle op to use new random generator.

* Refactored random_shuffle op helper.

* Fixed debug tests with random ops tests.

* Implement random_shuffle op cuda kernel helper and tests.

* - provide cuda scatter_update

Signed-off-by: Yurii <yurii@skymind.io>

* Implementation of random_shuffle for linear case with cuda kernels and tests.

* Implemented random_shuffle with cuda kernels. Final revision.

* - finally gruCell_bp is completed

Signed-off-by: Yurii <yurii@skymind.io>

* Dropout op cuda helper implementation.

* Implemented dropout_bp cuda helper.

* Implemented alpha_dropout_bp with cuda kernel helpers.

* Refactored helper.

* Implementation of suppresion helper with cuda kernels.

* - provide cpu code fot hsvToRgb, rgbToHsv, adjustHue

Signed-off-by: Yurii <yurii@skymind.io>

* Using sort by value method.

* Implementation of image.non_max_suppression op cuda-based helper.

* - correcting and testing adjust_hue, adjust_saturation cpu/cuda code

Signed-off-by: Yurii <yurii@skymind.io>

* Added cuda device prefixes to declarations.

* Implementation of hashcode op with cuda helper. Initital revision.

* rnn cu impl removed

Signed-off-by: raver119 <raver119@gmail.com>
2019-07-20 23:20:41 +10:00
Alex Black 68ea5f3688
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 21:34:34 +10:00
skymindops b5f0ec072f Eclipse Migration Initial Commit 2019-06-06 15:21:15 +03:00