Hi Everyone,
I recently tried deploying my pipeline, but ran into this issue -
“All outputs that share a cached LLM upstream must be placed in the same non-default job grouping. Cache: bf40c52c-327b-4df9-acd2-98a5600480f0”
If you produce multiple outputs downstream of the same LLM transform and each of those outputs is deployed into different job groupings, Foundry gets confused about how to share the cached LLM results.
To fix it, I went to the build settings and assigned all the outputs to the same custom job grouping. I was then able to deploy it successfully. Just wanted to highlight this as it was not mentioned in the Speedrun.
Hopefully this is helpful to someone else! Thanks!