coriander-dnn

Coriander-dnn provides a partial implementation of the NVIDIA® CUDA™ cuDNN API, for Coriander, OpenCL 1.2

Installation

• first, install coriander
• then run cocl_plugins install --repo-url https://github.com/hughperkins/coriander-dnn

Testing

Smoke test

This is mostly just to check plugins are working ok. Plugins are new :-)

Download https://github.com/hughperkins/coriander-dnn/raw/master/test/endtoend/basic1.cu to an empty folder somewhere, then, from that folder, do:

cocl_py --clang-home /usr/local/opt/llvm-4.0 basic1.cu
# hopefully compiles ok
# then run it
./basic.cu
# hopefully will print the model of your gpu at least

Unit tests

There are unit tests in test/gtest. You can build them:

make -j 8 tests

And run them:

make run-tests

cudnn test

This test uses the cudnn code at https://github.com/tbennun/cudnn-training to test that we can run convolutions and so on. I modified it slightly, to add a USE_OPENCL option, https://github.com/hughperkins/cudnn-training

To build cudnn-training using Coriander-dnn, you can do the following

• first install Coriander, and the Coriander-dnn plugin
• then build cudnn-training:

git clone https://github.com/hughperkins/cudnn-training
cd cudnn-training
mkdir build
cd build
ccmake ..
# press 'c' configure
# ignore the error about NVIDIA® CUDA™ toolkit not found, we dont need it
# change `USE_CUDA` to off
# change `USE_OPENCL` to on
# press 'c' configure, then 'g' generate
make

• download the mnist data:

wget http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz
wget http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz
wget http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz
wget http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz
gunzip train-images-idx3-ubyte.gz
gunzip train-labels-idx1-ubyte.gz
gunzip t10k-images-idx3-ubyte.gz
gunzip t10k-labels-idx1-ubyte.gz

• run:

./lenet

You should see iterations start running.