Hello! I am a research programmer for HuBMAP Consortium. We currently have our own cwltool fork for CUDA usage in our workflows, but we would like to use the standard cwltool version if possible. Basically, I need cwltool to say --gpus=device0:0
to Docker. Is that possible? How do I do that? Thank you!
Hello and welcome!
Please use the cwltool:CudaRequirement
extension and run cwltool
or toil-cwl-runner
using --enable-ext
.
Don’t forget to add the following somewhere in your CWL document:
$namespaces:
cwltool: "http://commonwl.org/cwltool#"