* - add causal mode of padding to convolutions
Signed-off-by: Yurii <iuriish@yahoo.com>
* - add additional tests for causal conv1d
Signed-off-by: Yurii <iuriish@yahoo.com>
* - add causal mode for cuda conv kernels
Signed-off-by: Yurii <iuriish@yahoo.com>
* Java side of Conv1D changes
Signed-off-by: raver119 <raver119@gmail.com>
* Add Conv1DDerivative op
Signed-off-by: Alex Black <blacka101@gmail.com>
* Causal Conv1D gradient checks
Signed-off-by: Alex Black <blacka101@gmail.com>
* Tweaks
Signed-off-by: Alex Black <blacka101@gmail.com>
* - add causal padding mode to conv2d_bp
Signed-off-by: Yurii <iuriish@yahoo.com>
* More thorough causal conv1d tests
Signed-off-by: Alex Black <blacka101@gmail.com>
* Implementation for non_max_suppression_v3 was added. Initial version
* Added check for overcome threshold.
* Added definition for V3 method.
* java remapping for NonMaxSuppressionV3
Signed-off-by: raver119 <raver119@gmail.com>
* Fixed proporly processing of an empty output and test.
* Refactored op to less threshold data to float.
* Implemented cuda-based helper for non_max_suppression_v3 op.
* Fixed fake_quant_with_min_max_vars op.
* Fixed tests with float numbers.
* - assert now stops execution
- sortByKey/sortByValue now have input validation
Signed-off-by: raver119 <raver119@gmail.com>
* missing var
Signed-off-by: raver119 <raver119@gmail.com>
* Fixed proper processing for zero max_size inputs.
* Refactored kernel callers.
* Fixed return statement for logdet op helper.
* Refactored unsorted segment SqrtN op.
* get back 8 tail bytes on CUDA
Signed-off-by: raver119 <raver119@gmail.com>
* Refactored segment prod ops and helpers for cuda and tests.
* Additional test.
* CudaWorkspace tests updated for 8 tail bytes
Signed-off-by: raver119 <raver119@gmail.com>
* special atomic test
Signed-off-by: raver119 <raver119@gmail.com>
* atomicMul/atomicDiv fix for 16bit values
Signed-off-by: raver119 <raver119@gmail.com>
* Eliminated waste prints.
* - improve performance of scatter (no lock) ops for 1D case
Signed-off-by: Yurii <iuriish@yahoo.com>
* - improve scatter lock op performance for 1D case
Signed-off-by: Yurii <iuriish@yahoo.com>
* - add kernel for verification of input indices-array elements in scatter and scatter_nd ops
Signed-off-by: Yurii <iuriish@yahoo.com>
* - provide fast indices checking on cpu side for scatter and gather osp
Signed-off-by: Yurii <iuriish@yahoo.com>
* - apply corrections requested by pr reviewer
Signed-off-by: Yurii <iuriish@yahoo.com>
* Corrected input checking and tests for bitcast op.
* Fixed an issue with non_max_suppression form generation and processing with score threshold given.
* Fixed bilinear resize kernel and tests.
* push for Serhii
Signed-off-by: raver119 <raver119@gmail.com>
* Added test for nearest_neighbor resize with int input.
* Added data type check for input/output match.
* Eliminate error in macros.
* Improved output message for type checking.
* Fixed input/output types for op.
* Eliminated waste logging.
* Refactored resize_bilinear helper for multithreading for cpu platform.
* Cosmetic changes only.
* Fixed error for string substitution.
* Skip test for cbow_batch with cuda.
* fix for resizeNearestNeighbor output dtype
Signed-off-by: raver119 <raver119@gmail.com>
* Refactored non_max_suppression helper.
* Refactored shape generation and input handling.
* Added additional test.
* - profiling cuda kernels for vol2col and im2col
Signed-off-by: Yurii <iuriish@yahoo.com>
* - correct addBias helper
Signed-off-by: Yurii <iuriish@yahoo.com>
* - correct mkl dilation formula and switch off mkl api for dilation deconvolutions
Signed-off-by: Yurii <iuriish@yahoo.com>
* Added implementation files for image_resize and resize_bicubic ops.
* Image resize and image.resize_bicubic ops implementation. Initial revision.
* Finished with infrastructure development for image.resize_bilinear op and image_resizo op implementation.
* Refactored resize methods.
* Added processing for Mitchelcubic algorithm.
* Added check for input/output sizes.
* Added int and float types for crop_and_resize op.
* Refactored crop_and_resize output type check.
* Added helper for bicubic interpolation as TF v.1 does.
* Added TF v.1 bicubic helper for cuda platform.
* Added cached class for bicubic algorithm.
* Refactored cuda implementation for crop_and_resize helper to use proper output type.
* Added facilities for bicubic interpolation.
* Portion bicubic interpolation from TF.
* Added tests for resize_bilinear testing.
* Working implementation of bicubic interpolation and tests.
* Refactored routines with image_resize bicubic op helper.
* Refactored code with coding standards.
* Refactored cpu helpers for resize_bicubic op.
* Refactored bicubic helpers.
* Added bicubic resize facilities.
* Implementing cuda kernels for bicubic interpolation. Implementation step.
* Cuda implementation of resize_bicubic op helper.
* Refactor image.resize_bicubic op helpers.
* Refactored helpers for resize_bicubic. Added error checking with cuda implementation.
* Refactored cuda implementation of resize_bicubic op helper. The first working revision.
* Cuda arch implementation for resize_bicubic op helper. Full working single-threaded revision.
* Intermediate bicubic interpolation helper for cuda.
* Refactored cpu helper for resize_bicubic.
* Multithreaded cuda implementation for resize_bicubic.
* Fixed merge issues.
* Refactored nlp helpers.
* Replicated resize_bicubic for 3D also.
* Eliminated waste comments of unused code.
* Eliminated waste comments with unused code.
* Eliminated waste template definitions.
* Eliminated waste debug code.
* Eliminated waste comments.
* Fixed multithreading with helpers.
* Fixed test suites for float and double in float point input lists.
* Fixed usage of reshape with 3D/4D on resizes.
* Final fixes.
* Fixed resize_neighbor op problem.
* Added a pair of tests for failed ops.
* Fixed cpu helper for draw_bounding_boxes op.
* Refactored implementation of draw_bounding_boxes op to full conform with TF.
* Improved multithreading with draw_bounding_boxes op cuda helper.
* Eliminated log messages.
* Changed logging with draw_bounding_boxes op helper and tests.
* Resize_biliear with 3D input allowed.
* Refactored 3D input acception with resize_bilinear op.
* And another improvement.
* Refactored reshape of input/output for resize_bilinear.
* Improvements final.
* Finished with 3D replication for image.resize_bilinear/_nearest_neighbor.
* Added copyrights for TF code.
* Using new form of multithreading for cpu implementation.
* Fixed shape error.
* Added multithreaded with batches on crop_and_resize functor.
* Refactored multithreading with crop_and_resize and draw_bounding_boxes.
* - get rid of some copy procedures in mmulHelper ops
Signed-off-by: Yurii <iuriish@yahoo.com>
* - further work on embedding cuda api for batched gemm (cublasGemmBatchedEx) in our mmulHelper class
Signed-off-by: Yurii <iuriish@yahoo.com>
* - further work on cuda batched gamm api
Signed-off-by: Yurii <iuriish@yahoo.com>
* - write own cuda kernel performing batched gemm
Signed-off-by: Yurii <iuriish@yahoo.com>
* missing include in MmulHelper
Signed-off-by: raver119 <raver119@gmail.com>
* - forgot to keep in code previous correct kernels for mmulNxN, since it may happen that new onw will fail for some reason in future
Signed-off-by: Yurii <iuriish@yahoo.com>
* disable old tensordot
Signed-off-by: raver119 <raver119@gmail.com>
* - rewrite cuda kernels for usualGemm and usualGemv
Signed-off-by: Yurii <iuriish@yahoo.com>
* - profiling mmul helpers
Signed-off-by: Yurii <iuriish@yahoo.com>
* - prints to check shapes were added
Signed-off-by: Yurii <iuriish@yahoo.com>
* - correct type of output array Cin mmulNxN
Signed-off-by: Yurii <iuriish@yahoo.com>
* - take into account possible nans in C array
Signed-off-by: Yurii <iuriish@yahoo.com>
* slightly change numThreads message
Signed-off-by: raver119 <raver119@gmail.com>
* - make corrections in accordance to given notes in pr review
Signed-off-by: Yurii <iuriish@yahoo.com>
* one range test
Signed-off-by: raver119 <raver119@gmail.com>
* few Context convenience singatures
Signed-off-by: raver119 <raver119@gmail.com>
* one more range test
Signed-off-by: raver119 <raver119@gmail.com>
* "range" "fix"
Signed-off-by: raver119 <raver119@gmail.com>
* adjuct_contrast_v2 now allows scale factor to be provided via input_variable
Signed-off-by: raver119 <raver119@gmail.com>
* adjust_contrast now allows scale factor as variable too
Signed-off-by: raver119 <raver119@gmail.com>
* bitcast shape tests
Signed-off-by: raver119 <raver119@gmail.com>
* BitCast import dtype added
Signed-off-by: raver119 <raver119@gmail.com>
* few more BitCast signatures
Signed-off-by: raver119 <raver119@gmail.com>
* Corrected randomuniform declaration.
* Refactored uniform distribution for both cuda and cpu platforms.
* Refactored uniform distribution and tests.
* Fixed type usage with indices.
* Refactored uniform distribution implementation and tests to full conform with TF implementation.
* Refactored gamma function to use type util method.
* Copyright changes and fixes with ConstantHelper.
* Added error checking on allocate cuda device memory and operations.
* Added implementation for random_gamma op.
* Added implementation for random_poisson op and support classes.
* Added helpers for random_poisson and random_gamma ops.
* Implementation of random_poisson. The first working edition.
* Implementation of random_poisson. Parallelized working edition.
* Implementation of random_gamma. Parallelized working edition with alpha only.
* Added cuda implementation for helper of poisson distribution.
* Corrected shape calculation with random_gamma and tests.
* Finished cpu implementation for gamma distribution.
* Finished cuda implementation for random_gamma op.
* Refactored cpu helpers for random_gamma and random_poisson ops.
* Refactored cuda helpers for gamma and poisson distribution.
* Refactored cuda helper for gamma distribution.
* Refactored cpu helper for random_poisson op.
* Refactored cpu helper for random_gamma op.
* - provide possibility to pass axis as last input array in concat op
- corrcect sumation in bias_add_bp op for NHWC case
Signed-off-by: Yurii <iuriish@yahoo.com>
* - write code for deconv2d op based on mkl dnn api
* no unsafe math
Signed-off-by: raver119 <raver119@gmail.com>
* no unsafe math
Signed-off-by: raver119 <raver119@gmail.com>
* - get rid of e<> and p<> methods in svd helper
Signed-off-by: Yurii <iuriish@yahoo.com>
* - provide mkl api support for deconvolution 3d
Signed-off-by: Yurii <iuriish@yahoo.com>
* - write deconv2d_bp based on mkl api
Signed-off-by: Yurii <iuriish@yahoo.com>
* - write deconv3d_bp based on mkl api
Signed-off-by: Yurii <iuriish@yahoo.com>
* - testing and fixing deconv based on mkl api
Signed-off-by: Yurii <iuriish@yahoo.com>
* - remove dilation form conv2d/3d mkl
Signed-off-by: Yurii <iuriish@yahoo.com>
* - minor changes
Signed-off-by: Yurii <iuriish@yahoo.com>
* - further corrections of deconv ops based on mkl dnn api
Signed-off-by: Yurii <iuriish@yahoo.com>
* - provide deconv2d_tf based on mkl dnn api
Signed-off-by: Yurii <iuriish@yahoo.com>
* - add minor corrections required by reviewer
Signed-off-by: Yurii <iuriish@yahoo.com>
* Added non_max_suppression_overlaps op and tests.
* Refactored implementation of non_max_suppression_overlaps.
* Refactoring of implementation of non_max_suppression_overlaps op.
* Refactoring of implementation of non_max_suppression op.
* Fixed portion error.
* Added cuda frontends for image suppression ops.
* Eliminated crash with cuda arch on image.non_max_suppression_overlaps op.
* Improved implementation of image_suppression helper for cpu platform.
* The generic approach of non_max_suppression_overlaps op helper with cuda platform.
* Working cuda implementation of helper non_max_suppression_overlaps op.
* Eliminated waste comments.
* Improved implementations for both platforms
* Refactored cuda implementation of image.non_max_suppression_overlaps op helper.
* Improved cuda implementation of non_max_suppression op helper.
* Refactored cuda implementation of image.non_max_suppression_overlaps op helper.
* Improved cuda implementation of image.non_max_suppression_overlaps op helper.
* Added modifications into cuda implementation for image suppression overlaps op.
* Correct queue emulation with cuda implementation of non_max_suppression_overlaps op.
* Prefinal stage of cuda implementation of non_max_suppression_overlaps.
* Worked cuda implementation of non_max_suppresion_overlaps helper.
* Fixed return to proper thread.
* Improvements for cuda implementation of image.non_max_suppression_overlaps op helper.
* Fixed implementation issues with non_max_suppression_overlaps on cuda platform.
* Fixed skip for non_max_suppression_overlaps on cuda platform.
* Finalize implementation of image_suppression helper and tests.
* Cosmetic changes only.
* - write code for new batchnorm backprop
Signed-off-by: Yurii <iuriish@yahoo.com>
* - testing batchnorm backprop
Signed-off-by: Yurii <iuriish@yahoo.com>
* - write code for batchnorm backprop based on mkl dnn api
Signed-off-by: Yurii <iuriish@yahoo.com>
* - testing and fixing bugs in batchnorm_bp mkl dnn
Signed-off-by: Yurii <iuriish@yahoo.com>
* - made corrections required by reviewer
Signed-off-by: Yurii <iuriish@yahoo.com>
* - change name in java wrapper for batchnorm op
Signed-off-by: Yurii <iuriish@yahoo.com>
shugeo
Yurii Shyrma
raver119
Alex Black
Alexander Stoyakin