Hi Garrett, thanks for promptly looking into this! I think i can live with this for a while, I managed to workout using a trick like this:
op:
...
op_looped:
... # same as op
requires:
- operation: op
select: [Saved_model]
train-recipe:
steps:
- op [flags 1]
- op_looped [flags 2]
- op_looped [flags 3]
Thanks for your insights!