Surfaced during bot review of #518.
Context
The three speech-recognition launch templates handle `container_config.revision` inconsistently:
- speech_recognition_slurm.sh always emits `--revision {{ container_config.revision }}`, even when the value is empty/None.
- speech_recognition_local.sh wraps it in `{%- if container_config.revision %}` on both docker and apptainer branches, so an unset revision is simply omitted.
If `container_config.revision` is ever unset on the Slurm path, the rendered command would include `--revision` with an empty value. SRI 0.2.1's typer CLI accepts `--revision` as `Optional[str]`, so an empty string may or may not do the right thing depending on how the model store resolves it — worth confirming, but at minimum the Slurm template should behave like the local one.
Scope
Surfaced during bot review of #518.
Context
The three speech-recognition launch templates handle `container_config.revision` inconsistently:
If `container_config.revision` is ever unset on the Slurm path, the rendered command would include `--revision` with an empty value. SRI 0.2.1's typer CLI accepts `--revision` as `Optional[str]`, so an empty string may or may not do the right thing depending on how the model store resolves it — worth confirming, but at minimum the Slurm template should behave like the local one.
Scope