Hi there! We have some flows that need to multiprocessing for ML workloads, which require additional shared memory available under

/dev/shm

or use the host’s namespace (setting --ipc=‘host’ or

shareProcessNamespace=true

at runtime). We saw there was an open issue trying to solve this problem, but it looks like it was dropped: https://github.com/PrefectHQ/prefect/issues/12924 Is there a known solution for this? Ideally by modifying the base job template for our work pool / adding a kwarg in our

prefect.yaml

when defining our deployment config, etc.