Nnabla-ext-cuda

release-note-bugfix

+ [Fix Gelu Backward](https://github.com/sony/nnabla-ext-cuda/pull/235)
+ [Make scatter_nd work in static graph.](https://github.com/sony/nnabla-ext-cuda/pull/234)

release-note-build

+ [Build Multi GPU option with Specified MPI version.](https://github.com/sony/nnabla-ext-cuda/pull/229)

release-note-doc

+ [Add licenses and Notice ](https://github.com/sony/nnabla-ext-cuda/pull/230)

release-note-op-layer

+ [SpatialTransformer](https://github.com/sony/nnabla-ext-cuda/pull/233)

release-note-utility

+ [Add DLPack interface to share a tensor on host or device among frameworks and libraries without copy](https://github.com/sony/nnabla-ext-cuda/pull/232)

Install the latest nnabla by:

pip install nnabla
pip install nnabla-ext-cuda102 For CUDA version 10.2 users


You can also install the cuda extension with specific versions from one of the following. [See also FAQ](https://nnabla.readthedocs.io/en/latest/python/install_on_linux.html#faq)

release-note-bugfix
+ [[fix][c++] Allow calling multiple times and catch errors in FFT/IFFT](https://github.com/sony/nnabla-ext-cuda/pull/227)
+ [change data type of size from int to int64 to fix overflow.](https://github.com/sony/nnabla-ext-cuda/pull/221)

release-note-build
+ [update devtoolset package version. update url of yum repository](https://github.com/sony/nnabla-ext-cuda/pull/222)

release-note-doc
+ [Fix brand name](https://github.com/sony/nnabla-ext-cuda/pull/228)

release-note-op-layer
+ [Quantize/DequantizeLinear](https://github.com/sony/nnabla-ext-cuda/pull/226)
+ [Add deconvolution output padding option.](https://github.com/sony/nnabla-ext-cuda/pull/224)
+ [Interpolation with half_pixel and half_pixel_for_nn](https://github.com/sony/nnabla-ext-cuda/pull/223)

release-note-utility
+ [Add a flag to choose conv algo by a heuristic](https://github.com/sony/nnabla-ext-cuda/pull/225)

Install the latest nnabla by:

pip install nnabla
pip install nnabla-ext-cuda102 For CUDA version 10.2 users


You can also install the cuda extension with specific versions from one of the following. [See also FAQ](https://nnabla.readthedocs.io/en/latest/python/install_on_linux.html#faq)

release-note-core

+ [Resolve some compiler warnings](https://github.com/sony/nnabla-ext-cuda/pull/219)

release-note-op-layer

+ [Unpooling with Channel Last](https://github.com/sony/nnabla-ext-cuda/pull/220)

release-note-utility

+ [Add NanInfTracer ](https://github.com/sony/nnabla-ext-cuda/pull/218)

Install the latest nnabla by:

pip install nnabla
pip install nnabla-ext-cuda101 For CUDA version 10.1 users


Note that CUDA 10.1 and cuDNN 7.6 are set as default if versions are not specified. You can also install the cuda extension with specific versions from one of the following. [See also FAQ](https://nnabla.readthedocs.io/en/latest/python/install_on_linux.html#faq)

release-note-bugfix
+ [Fix type name is not allowed.](https://github.com/sony/nnabla-ext-cuda/pull/216)

release-note-build
+ [Fix CPU Library name for creating wheel.](https://github.com/sony/nnabla-ext-cuda/pull/212)

release-note-op-layer
+ [Double Backward for ChannelLast and Half Type](https://github.com/sony/nnabla-ext-cuda/pull/217)
+ [RandomErase Function](https://github.com/sony/nnabla-ext-cuda/pull/214)

release-note-utility
+ [FIX: Compatibility with DALI 0.18.0](https://github.com/sony/nnabla-ext-cuda/pull/213)

Install the latest nnabla by:

pip install nnabla
pip install nnabla-ext-cuda101 For CUDA version 10.1 users


Note that CUDA 10.1 and cuDNN 7.6 are set as default if versions are not specified. You can also install the cuda extension with specific versions from one of the following. [See also FAQ](https://nnabla.readthedocs.io/en/latest/python/install_on_linux.html#faq)

release-note-bugfix

+ [Support a deterministic algorithm at cudnnMaxPooling](https://github.com/sony/nnabla-ext-cuda/pull/211)
+ [Properly handle MPI errors](https://github.com/sony/nnabla-ext-cuda/pull/209)

release-note-build

+ [Fix CPU Library name for creating wheel.](https://github.com/sony/nnabla-ext-cuda/pull/212)
+ [Enable building CLI with CUDA](https://github.com/sony/nnabla-ext-cuda/pull/206)

release-note-op-layer

+ [Add AddN and MulN functions](https://github.com/sony/nnabla-ext-cuda/pull/210)
+ [Add RandomCropCuda](https://github.com/sony/nnabla-ext-cuda/pull/208)
+ [Add CUDA implementation of patch_correlation function.](https://github.com/sony/nnabla-ext-cuda/pull/205)
+ [CUDNN NHWC deconvolution](https://github.com/sony/nnabla-ext-cuda/pull/204)
+ [Adaptive Separable Convolution](https://github.com/sony/nnabla-ext-cuda/pull/203)


Install the latest nnabla by:


pip install nnabla
pip install nnabla-ext-cuda101 For CUDA version 10.1 users


Note that CUDA 10.1 and cuDNN 7.6 are set as default if versions are not specified. You can also install the cuda extension with specific versions from one of the following. [See also FAQ](https://nnabla.readthedocs.io/en/latest/python/install_on_linux.html#faq)

release-note-bugfix

+ [[Windows] Fix batchnormalization fail with CUDNN_STATUS_NOT_SUPPORTED](https://github.com/sony/nnabla-ext-cuda/pull/199)

release-note-build

+ [Remove spdlog and python2 dependency.](https://github.com/sony/nnabla-ext-cuda/pull/202)
+ [Remove python2 dependencies.](https://github.com/sony/nnabla-ext-cuda/pull/200)
+ [Remove binary-only option from pip in Dockerfile.](https://github.com/sony/nnabla-ext-cuda/pull/196)

release-note-op-layer

+ [Add CUDA implementation for warp_by_flow function.](https://github.com/sony/nnabla-ext-cuda/pull/195)
+ [Implement batch_det](https://github.com/sony/nnabla-ext-cuda/pull/194)
+ [BatchNorm backward in test mode](https://github.com/sony/nnabla-ext-cuda/pull/193)
+ [Add batchinv function](https://github.com/sony/nnabla-ext-cuda/pull/192)

release-note-utility

+ [Implement clip_grad_by_norm at solvers](https://github.com/sony/nnabla-ext-cuda/pull/197)
+ [Support any dimensional array as an input of pooling functions](https://github.com/sony/nnabla-ext-cuda/pull/198)


Install the latest nnabla by:


pip install nnabla
pip install nnabla-ext-cuda101 For CUDA version 10.1 users


Note that CUDA 10.1 and cuDNN 7.6 are set as default if versions are not specified. You can also install the cuda extension with specific versions from one of the following. [See also FAQ](https://nnabla.readthedocs.io/en/latest/python/install_on_linux.html#faq)