This example multiplies matrix a by matrix b, producing a * b.
Both matrices, a and b, must be the same type.
Matrix dimensions for matrix a and matrix b are specified by the user in file ./workspace/shapes.yaml
The matrix dimensions must be >=2 and the 2 inner dimensions must specify valid matrix multiplication dimensions.
virtualenv -p python3 ./env && source ./env/bin/activate
pip install mlcube mlcube-docker mlcube-singularity mlcube-ssh
git clone https://github.com/mlperf/mlcube_examples.git && cd ./mlcube_examples/matmul
# Configure Matmul MLCube
mlcube_docker configure --mlcube=. --platform=platforms/docker.yaml
# Run Matmul tasks: You can change the shapes of the matrices that are multipled in workspace/shapes.yaml
mlcube_docker run --mlcube=. --platform=platforms/docker.yaml --task=run/matmul.yaml
Go to workspace/
directory and study its content. Then:
ls ./workspace
cat ./workspace/matmul.txt