Toil 8.x prints the output of nvidia-smi to stderr at the beginning of each run
Is this a forgotten debugging statement? It can be kind of confusing and annoying.
Ex: Someone just installed Cactus and ran cactus --help and the first thing that appeared was:
Failed to initialize NVML: Driver/library version mismatch
NVML library version: 550.144
and contacted me about their installation being broken. But they only want to run the cpu-only pangenome pipeline so this error doesn't matter. On the flipside, even if their driver was setup okay, they probably don't want to see the full chart output of nvidia-smi when they run cactus --help either.
┆Issue is synchronized with this Jira Story ┆Issue Number: TOIL-1751
It looks like this is because we use check_call and the default of printing stdout directly in the call: https://github.com/DataBiosphere/toil/blob/a20dc7e59b722a2282a3f8cb8b811488e0e56edb/src/toil/lib/accelerators.py#L37
I'll make a PR to fix this. Thanks for reporting it.
PR is here: https://github.com/DataBiosphere/toil/pull/5307