* Fix MergeVertex serialization for NHWC case
Signed-off-by: Alex Black <blacka101@gmail.com>
* #8999 Dropout JSON field ignore
Signed-off-by: Alex Black <blacka101@gmail.com>
* Added dtype formulation for poisson and gamma distributions.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Refactored gamma distribution generator and tests.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Added generator for gamma distribution when alpha (shape) between 0 and 1
Signed-off-by: shugeo <sgazeos@gmail.com>
* Implemented gamma distribution for shape param less than 1 and tests.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Implemented gamma distributed randoms for shape (alpha) parameter greater then 1.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Added cuda implementation for gamma distribution.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Refactored cuda and cpu implementation of gamma distribution.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Fixed crash with default beta param with gamma distribution.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Fixed pow for arm arch.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Gamma test fixed
* Cosmetic changes only.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Fixed random value retrieving
* Eliminated overflow attemptions.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Modified random retrieving.
Signed-off-by: shugeo <sgazeos@gmail.com>
* enlighted density of tests for Gamma distribution.
Signed-off-by: shugeo <sgazeos@gmail.com>
Co-authored-by: Alexander Stoyakin <alexander.stoyakin@gmail.com>
Co-authored-by: raver119 <raver119@gmail.com>
* get rid of culibos reference
Signed-off-by: raver119 <raver119@gmail.com>
* typo
Signed-off-by: raver119 <raver119@gmail.com>
* one less printf
Signed-off-by: raver119 <raver119@gmail.com>
* one disabled test
Signed-off-by: raver119 <raver119@gmail.com>
* Fix Yolo2OutputLayer for NHWC data format
Signed-off-by: Alex Black <blacka101@gmail.com>
* ObjectDetectionRecordReader NHWC support
Signed-off-by: Alex Black <blacka101@gmail.com>
* Eliminated error with resize implementation.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Refactored resize caller implementation.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Refactored image.resize op helper.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Added dumb implementations for missed resize methods.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Added resize_images op. Refactored image_resize op.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Refactored matrix_band_part op and test.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Refactored resize_images op to comply with preserve_aspect_ratio flag properly.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Refactored resize_images and tests for resizeArea method.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Refactored resize methods and test.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Added new methods for TF2 resize op.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Portion of resize algorithms from TF2
Signed-off-by: shugeo <sgazeos@gmail.com>
* Added routine to process resize with given algorithm.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Added new image resize via scale and translate process helper.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Cpu implementation for V2 image resize operation helpers.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Added implementation for lancos5 algorithm of resize and test.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Added prints for span computing.
Signed-off-by: shugeo <sgazeos@gmail.com>
* The first working implementation and tests for lancos5 resize.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Eliminated waste prints.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Refactored image_resize op and tests."
Signed-off-by: shugeo <sgazeos@gmail.com>
* Lanczos3 resize implementation and tests.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Implemented bicubic resize algorithm and tests.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Added a couple of tests and cosmetic changes with image resize helper.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Added bilinear implementation for image resize.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Refactored bicubic algorithm and also implement area and neighbor algoritms for image resize on cpu arch.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Added a couple of tests for nearest neighbor and area resize.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Cosmetic changes for cpu implementation and added cuda implementation for resize methods.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Separated cuda implementation of v2 image resize.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Added kernels for span calculation and span gathering with new image resize cuda implementation.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Refactored cuda implementation of image resize kernels.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Finished the first working implementation of image resize op and tests.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Fixed resize_images and image_resize ops.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Refactored shape construction and output validation.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Fixed test to properly initalized with float.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Added 3D input opotunity for resize ops.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Fixed test for resize_images op.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Fixed test and call for resize_images op.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Refactored image_resize op output data type handling for nearest neighbors method and tests.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Fixed issue with wrong resize method.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Added checkup for wrong resize methods for resize ops.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Refactored resize methods and test.
Signed-off-by: shugeo <sgazeos@gmail.com>
* Added output data type validation for given resize method.
Signed-off-by: shugeo <sgazeos@gmail.com>
* - ResizeMethod rearranged in order to match C++ side
- minor test fix
Signed-off-by: raver119@gmail.com <raver119@gmail.com>
* Refactored resize_images op.
Signed-off-by: shugeo <sgazeos@gmail.com>
Co-authored-by: raver119@gmail.com <raver119@gmail.com>
* - disable mkldnn concat when number of input arrays > 3072
Signed-off-by: Yurii <iuriish@yahoo.com>
* - get rid of loop in calculating of input arrays number
Signed-off-by: Yurii <iuriish@yahoo.com>
* Add SameDiff file format ADR
Signed-off-by: Alex Black <blacka101@gmail.com>
* Update 0001-SameDiff_File_Format.md
* Update
Signed-off-by: Alex Black <blacka101@gmail.com>
* Timeouts and fixes
Signed-off-by: Alex Black <blacka101@gmail.com>
* Increase default timeout to 90s due to slow PPC CI machines
Signed-off-by: Alex Black <blacka101@gmail.com>
* Another timeout tweak
Signed-off-by: Alex Black <blacka101@gmail.com>
* Svhn
Signed-off-by: Alex Black <blacka101@gmail.com>
* - 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>
* error code check in CudaMemoryManager
Signed-off-by: raver119@gmail.com <raver119@gmail.com>
* clear
Signed-off-by: raver119@gmail.com <raver119@gmail.com>
* clear model before exiting
Signed-off-by: raver119@gmail.com <raver119@gmail.com>
* MultiLayerNetwork/ComputationGraph.close() [WIP] (#460)
* MultiLayerNetwork/ComputationGraph.close()
Signed-off-by: Alex Black <blacka101@gmail.com>
* Copyright header
Signed-off-by: Alex Black <blacka101@gmail.com>
* Fix
Signed-off-by: Alex Black <blacka101@gmail.com>
* - fix for handling release of nested DataBuffers
- couple of additional tests for released DataBuffers
Signed-off-by: raver119@gmail.com <raver119@gmail.com>
* PW test: increase number of epochs slightly
Signed-off-by: raver119@gmail.com <raver119@gmail.com>
Co-authored-by: Alex Black <blacka101@gmail.com>
* - numPrefixBlocks fix for threshold_encoding
- temparrays pointers fixed
Signed-off-by: raver119@gmail.com <raver119@gmail.com>
* auto configuration of memory workspace for gradients sharing
Signed-off-by: raver119@gmail.com <raver119@gmail.com>
* limit sparse encoding message size
Signed-off-by: raver119@gmail.com <raver119@gmail.com>
* one more workspace test
Signed-off-by: raver119@gmail.com <raver119@gmail.com>
* one more CUDA-specific test
Signed-off-by: raver119@gmail.com <raver119@gmail.com>
* one more CUDA-specific workspace test
Signed-off-by: raver119@gmail.com <raver119@gmail.com>
* one more CUDA-specific workspace test
Signed-off-by: raver119@gmail.com <raver119@gmail.com>
* one more CUDA-specific workspace test
Signed-off-by: raver119@gmail.com <raver119@gmail.com>
* add separate host/device reset for circular workspace mode
Signed-off-by: raver119@gmail.com <raver119@gmail.com>
* new PW builder method for encoder memory amount
Signed-off-by: raver119@gmail.com <raver119@gmail.com>
* "inplace" execution for threshold encoding
Signed-off-by: raver119@gmail.com <raver119@gmail.com>
* Remove debug line from KerasConvolution2D
Signed-off-by: Alex Black <blacka101@gmail.com>
* Remove more debug lines
Signed-off-by: Alex Black <blacka101@gmail.com>
* - provide correct possible output types in mergeMaxIndex op
Signed-off-by: Yurii <iuriish@yahoo.com>
* - cleaning up the unneeded backprop arg in reverse_bp op
Signed-off-by: Yurii <iuriish@yahoo.com>
* - improve clipByNorm both ff and bp
Signed-off-by: Yurii <iuriish@yahoo.com>
* - implementation and testing clipByAvgNorm_bp op
Signed-off-by: Yurii <iuriish@yahoo.com>
* - pass biases in any way in dnnl lstm op, they are zeros when user doesn't provide them to us
Signed-off-by: Yurii <iuriish@yahoo.com>
* - start working on mkldnn concat op
Signed-off-by: Yurii <iuriish@yahoo.com>
* - further work on mkldnn concat
Signed-off-by: Yurii <iuriish@yahoo.com>
* missing declaration fix
Signed-off-by: raver119@gmail.com <raver119@gmail.com>
* - polishing mkl ops
Signed-off-by: Yurii <iuriish@yahoo.com>
* - testing and fixing bugs in mkl concat op
Signed-off-by: Yurii <iuriish@yahoo.com>
* - fix linkage error for windows cuda build
Signed-off-by: Yurii <iuriish@yahoo.com>
* - further conflicts resolving with master
Signed-off-by: Yurii <iuriish@yahoo.com>
* - fix format tags in mkldnn matmul op
Signed-off-by: Yurii <iuriish@yahoo.com>
* - provide additional type cast in clip.cu
Signed-off-by: Yurii <iuriish@yahoo.com>
* - finally bug in mkldnn tanh_bp was caught
Co-authored-by: raver119@gmail.com <raver119@gmail.com>
* Update docs links to new website URLs
Signed-off-by: Alex Black <blacka101@gmail.com>
* One more link
Signed-off-by: Alex Black <blacka101@gmail.com>
Alex Black
Paul Dubs
shugeo
raver119
Abdelrauf
Serhii Shepel
Alexandre Boulanger
Fariz Rahman
Yurii Shyrma
Andrii T