Open
Description
Migrated from mlcommons/mlperf-automations_archived#40
Originally created by @arjunsuresh on Mon, 26 Dec 2022 12:33:25 GMT
We can have data preprocessing scripts needing information about the model for which the dataset is going to be used but the model as such is not needed. It would be good to have a way to share the model meta between the model download script and the dataset preprocessing script which can avoid duplicating the meta information in _cm.json.
Can we add a dependency with no-deps and fake-run options so that only the script meta gets processed?