-
Notifications
You must be signed in to change notification settings - Fork 6.8k
[v1.x][BUGFIX] Implement oneDNN deconvolution primitives to deconvolution 2D #20107
Conversation
Hey @PawelGlomski-Intel , Thanks for submitting the PR
CI supported jobs: [windows-cpu, unix-gpu, website, unix-cpu, miscellaneous, sanity, edge, clang, windows-gpu, centos-gpu, centos-cpu] Note: |
@mxnet-bot run ci [unix-cpu, website] |
Jenkins CI successfully triggered : [website, unix-cpu] |
27eb164
to
cab1562
Compare
@mxnet-bot run ci [unix-gpu, centos-gpu] |
Jenkins CI successfully triggered : [centos-gpu, unix-gpu] |
@mxnet-bot run ci [unix-gpu] |
Jenkins CI successfully triggered : [unix-gpu] |
Merged. @PawelGlomski-Intel thank you! |
…tion 2D (apache#20107) * Use mkldnn deconvolution primitive in deconvolution * Apply clang-format * Refactor deconvolution version 1 * Refactor deconvolution version 2 and use permute_axes in IOLogicalSwapDesc * Refactor deconvolution version 3 * Enable Deconvolution2D test * Fix sanity * Fix windows builds * Fix deconvolution with bias test
…tion 2D (apache#20107) * Use mkldnn deconvolution primitive in deconvolution * Apply clang-format * Refactor deconvolution version 1 * Refactor deconvolution version 2 and use permute_axes in IOLogicalSwapDesc * Refactor deconvolution version 3 * Enable Deconvolution2D test * Fix sanity * Fix windows builds * Fix deconvolution with bias test
…tion 2D (apache#20107) * Use mkldnn deconvolution primitive in deconvolution * Apply clang-format * Refactor deconvolution version 1 * Refactor deconvolution version 2 and use permute_axes in IOLogicalSwapDesc * Refactor deconvolution version 3 * Enable Deconvolution2D test * Fix sanity * Fix windows builds * Fix deconvolution with bias test
…tion 2D (apache#20107) * Use mkldnn deconvolution primitive in deconvolution * Apply clang-format * Refactor deconvolution version 1 * Refactor deconvolution version 2 and use permute_axes in IOLogicalSwapDesc * Refactor deconvolution version 3 * Enable Deconvolution2D test * Fix sanity * Fix windows builds * Fix deconvolution with bias test
…t and fix bias (#20292) * [v1.x][BUGFIX] Implement oneDNN deconvolution primitives to deconvolution 2D (#20107) * Use mkldnn deconvolution primitive in deconvolution * Apply clang-format * Refactor deconvolution version 1 * Refactor deconvolution version 2 and use permute_axes in IOLogicalSwapDesc * Refactor deconvolution version 3 * Enable Deconvolution2D test * Fix sanity * Fix windows builds * Fix deconvolution with bias test * [v1.x][FEATURE] Add MKLDNN Deconvolution 1D and 3D support (#20137) * Use MXNET_USE_ONEDNN * Fix test * Apply formatter * Add native support for 3D deconvolution * Remove outdated check * Replace math.prod with np.prod * Check convolution layout only when it has value * Remove outdated check * Change tests * Increase default workspace size to mach convolution * Fix deconv workspace size * Increase default deconv workspace size in python API * Disable 3D tests for GPU * Add deconv arguments checks * Remove next_impl calls until it is fixed * Share workspace * Fix documentation * Add test_deconv_dilation * Fix check * Fix include order
Description
Currently, deconvolution 2D is implemented with oneDNN convolution primitives. Deconvolution forward pass is implemented with convolution backward primitives that do not include bias. This requires you to manually add the bias, which is currently broken (#19768).
This change implements oneDNN deconvolution primitives to deconvolution, fixing (#19768).
As I couldn't reproduce (#12579) for the 2D case, I enabled the 2D part of this test.
Checklist
Essentials
Changes