using inline config parameters (e.g., medperf --gpus=1 run) or setting config parameters (e.g., medperf profile set --platform=docker) don't have validation. So a user might use an invalid config parameter and may only know it's invalid after getting passed to mlcube.
It might be better to have checks on the MedPerf side
using inline config parameters (e.g.,
medperf --gpus=1 run) or setting config parameters (e.g.,medperf profile set --platform=docker) don't have validation. So a user might use an invalid config parameter and may only know it's invalid after getting passed tomlcube.It might be better to have checks on the MedPerf side