systemml git commit: [SYSTEMML-445] Avoid unnecessary transfer to the GPU for size estimation

Repository: systemml Updated Branches: refs/heads/master 3702df7c1 -> 97fd7d1aa [SYSTEMML-445] Avoid unnecessary transfer to the GPU for size estimation - Compute memory estimates (exact and worst-case) using metadata rather than requiring pointer transfer. Project: http://git-wip-us.apache.o

systemml git commit: [SYSTEMML-445] Improved the performance of batchnorm backward

Repository: systemml Updated Branches: refs/heads/master 512fb9e11 -> 3702df7c1 [SYSTEMML-445] Improved the performance of batchnorm backward - Added a custom kernel for computing dgamma in batch normalization layer. - Also, fixed a minor bug in GPUDenseInputPointerFetcher class. Project: htt

systemml git commit: [SYSTEMML-445] Extend coverage for GPU batchnorm test rewrite

Repository: systemml Updated Branches: refs/heads/master fab31fd1f -> 512fb9e11 [SYSTEMML-445] Extend coverage for GPU batchnorm test rewrite - If inv_var rewrite has already been applied, the application of GPU batchnorm test rewrite (and CuDNN batchnorm kernel) is skipped. This commit fixes

systemml git commit: [SYSTEMML-445] Fixed the error handling during GPU memory cleanup

Repository: systemml Updated Branches: refs/heads/master 8a144f2b3 -> fab31fd1f [SYSTEMML-445] Fixed the error handling during GPU memory cleanup If an error occurs during cleanup of temporary memory and free-ing of GPU context, SystemML does not display the correct error message. This commit

systemml git commit: [MINOR] Bugfix in Keras2DML API when loading weights from Keras

Repository: systemml Updated Branches: refs/heads/master ef15e582b -> 8a144f2b3 [MINOR] Bugfix in Keras2DML API when loading weights from Keras Project: http://git-wip-us.apache.org/repos/asf/systemml/repo Commit: http://git-wip-us.apache.org/repos/asf/systemml/commit/8a144f2b Tree: http://git