Version 1.32.0 Release

release-note-bugfix

• complete setup cfg in all condition
• Use not cast but get in nan/inf check
• Fix cython build failure by missing cutensor.h in non-docker env
• Not to all-reduce gradients of beta and gamma in sync BN
• Use a singleton function for cudaGetDeviceProperties to avoid slowdown

release-note-build

• add python310 into lib_in_wheel
• change pip installation method
• support python3.10
• Add ca-certificates to Dockerfiles

release-note-core

• Caching cuTENSOR handles and optimize TranposeCuda::setup_impl

release-note-op-layer

• Faster LayerNormCuda::setup_impl (GN and IN as well) by removing unnecessary function creation

release-note-utility

• add watchdog timeout env so that can set timeout

Install the latest nnabla by:

pip install nnabla
pip install nnabla-ext-cuda110 # For CUDA version 11.0.3 users

You can also install the cuda extension with specific versions from one of the following. See also FAQ

• nnabla-ext-cuda110 (CUDA 11.0.3 x cuDNN 8.0)
• nnabla-ext-cuda114 (CUDA 11.4.3 x cuDNN 8.0)

pip install nnabla
pip install nnabla_ext_cuda110  # For CUDA 11.0.3 x cuDNN 8.0 users

The nnabla cuda extension supports both centralized and distributed training.
Below packages detect MPI and NCCL2 installed on your system. If either one
of then is unavailable, distributed training is disabled.

• nnabla-ext-cuda110 (Ubuntu18.04 default)
• nnabla-ext-cuda114 (Ubuntu18.04 default)

Additional setup may be required depending on your OS or environment. Please check Python Package Installation Guide for details.

Install Guide
To use C++ inference feature, follow the demonstration on MNIST inference in C++.

Demonstration

Build Guide
The "nnabla-ext-cuda" package is temporarily unavailable. Use of this package is not recommended. Please install nnabla-ext-cuda110, nnabla-ext-cuda114 instead.
The following nnabla CUDA extension packages have been deprecated and the PyPi repository has been (or going to be) closed.

• nnabla-ubuntu16
• nnabla-ubuntu18
• nnabla-ext-cuda
• nnabla-ext-cuda80
• nnabla-ext-cuda91
• nnabla-ext-cuda92
• nnabla-ext-cuda100
• nnabla-ext-cuda101
• nnabla-ext-cuda90-nccl2-ubuntu16
• nnabla-ext-cuda100-nccl2-ubuntu16
• nnabla-ext-cuda100-nccl2-ubuntu18
  /home/jenkins/.local/lib/python3.6/site-packages/requests/init.py:104: RequestsDependencyWarning: urllib3 (1.26.12) or chardet (5.0.0)/charset_normalizer (2.0.12) doesn't match a supported version!
  RequestsDependencyWarning)
• nnabla-ext-cuda101-nccl2-ubuntu16
• nnabla-ext-cuda101-nccl2-ubuntu18
• nnabla-ext-cuda102-nccl2-mpi2-1-1
• nnabla-ext-cuda102-nccl2-mpi3-1-6
• nnabla-ext-cuda110-nccl2-mpi2-1-1
• nnabla-ext-cuda110-nccl2-mpi3-1-6

The following "nnabla-ext-cuda" docker images have been deprecated.

• py27-cuda92
• py36-cuda92
• py37-cuda92
• py27-cuda92-v1.0.xx
• py36-cuda92-v1.0.xx
• py37-cuda92-v1.0.xx
• py36-cuda100-v1.xx.x
• py37-cuda100-v1.xx.x
• py38-cuda100-v1.xx.x
• py39-cuda100-v1.xx.x
• py36-cuda102-v1.xx.x
• py37-cuda102-v1.xx.x
• py38-cuda102-v1.xx.x
• py39-cuda102-v1.xx.x

We've decided to change nnabla's versioning policy to semantic versioning.
This change has been applied from version 1.1.0.

Python 2 is no longer be supported from v1.5.0.
CUDA8.0 is no longer be supported from v1.6.0.
Python3.5 is no longer be supported from v1.14.0.
CUDA9.0 is no longer be supported from v1.14.0.
Python3.6 is no longer be supported from v1.26.0.
CUDA10.0 is no longer be supported from v1.26.0.
CUDA10.2 is no longer be supported from v1.30.0.

Known Issue:

• cufftXtMakePlanMany returns CUFFT_INVALID_PLAN on Windows + CUDA11.4.3 environment. See details here