Fix for multigpu runs with docker executor (#134)

* Fix for multigpu runs with docker executor

Signed-off-by: Igor Gitman <[email protected]>

* Style fix

Signed-off-by: Igor Gitman <[email protected]>

---------

Signed-off-by: Igor Gitman <[email protected]>

Author: Kipok

src/nemo_run/run/torchx_backend/schedulers/docker.py

@@ -88,7 +88,7 @@ def _submit_dryrun(self, app: AppDef, cfg: Executor) -> AppDryRunInfo[DockerJobR
             cmd = [role.entrypoint] + role.args
             containers.append(
                 DockerContainer(
-                    name=f"{executor.experiment_id}_{role.name}",
+                    name=f"{executor.experiment_id}_{role.name}".replace("_", "-"),
                     command=cmd,
                     executor=_current_executor,
                     extra_env=role.env,