NNabla CUDA extension package installation using PIP¶
Note: please refer to the OS specific workflows for the OS specific dependencies setup.
By installing the NNabla CUDA extension package nnabla-ext-cuda, you can accelerate the computation by NVidia CUDA GPU (CUDA must be setup on your environment accordingly).
Several pip packages of NNabla CUDA extension are provided for each CUDA version and its corresponding CUDNN version as following.
CUDA vs CUDNN Compatibility¶
| Package name | CUDA version | CUDNN version |
|---|---|---|
| nnabla-ext-cuda80 | 8.0 | 7.1 |
| nnabla-ext-cuda90 | 9.0 | 7.3(win), 7.4(linux) |
| nnabla-ext-cuda92 | 9.2 | 7.3(win), 7.4(linux) |
| nnabla-ext-cuda100 | 10.0 | 7.3(win), 7.4(linux) |
The latest CUDA version is always preferred if your GPU accepts.
Installation¶
The following is an example of installing the extension for CUDA 9.2.
pip install nnabla_ext_cuda92
and check if all works.
python -c "import nnabla_ext.cuda, nnabla_ext.cudnn"
2018-06-26 15:20:36,085 [nnabla][INFO]: Initializing CPU extension...
2018-06-26 15:20:36,257 [nnabla][INFO]: Initializing CUDA extension...
2018-06-26 15:20:36,257 [nnabla][INFO]: Initializing cuDNN extension...
Note: If you want to make sure the latest version will be installed, try to uninstall previously installed one with pip uninstall -y nnabla nnabla_ext_cuda92 beforehand.
Installation with Multi-GPU supported¶
Multi-GPU wheel package is available only on ubuntu16.04 and python3.5+, you can install as the following.
pip install nnabla-ubuntu16
pip install nnabla-ext-cuda92-nccl2-ubuntu16
If you already installed NNabla, uninstall all of it, or start from a clean environment which you create using Anaconda, virtualenv, or pyenv.
You should also install OpenMPI,
apt-get install libopenmpi-dev
and NCCL in addition to CUDA and CuDNN.
FAQ¶
Q. How do I install CUDA?¶
NNabla CUDA extension requires both CUDA toolkit and CUDNN library. You should select a proper CUDA version according to your CUDA device capability. See the official installation guide. NNabla supports CUDA versions later than 8.0. See the table for the CUDNN compatibility with the specific CUDA versions.
Q. How do I check proper version of cuDNN¶
Enter the following command:
python -c "import nnabla_ext.cuda, nnabla_ext.cudnn"
If there is a version mismatch on your machine, you can see proper versions in the error message. Following is a sample error message.
[nnabla][INFO]: Initializing CPU extension...
Please install CUDA version 9.2.
and CUDNN version 7.3.1.
Or install correct nnabla_ext_cuda for installed version of CUDA/CUDNN.