Like the Python executable and torchrun, the -m flag allows a CLI to call an installed Python module.
We should integrate the -m flag to torchrun-hpc by:
- Adding a flag to the
torchrun-hpc CLI: hpc_launcher/cli/torchrun_hpc.py
- Enabling module execution in the torchrun-hpc trampoline:
hpc_launcher/torch/torchrun_hpc_trampoline.py
We should also add a test.