4/20 healthy workers - pruna ai image model from hub
Hallo,
I am really stack with that, raised limits to 20 workers (even 30 in theory), I get 4 healthy workers, so queue of 100 images is done by 4 workers at most.
model from hub: FLUX.1-dev-juiced 0.1.2
obserservations:
1. deleted unhealthy - they were recreated and also unhealthy
2. A40 was always i a list but found 1 4090 healthy as well - so it semes do not relate to GPU
Cold boot times are x3 now comparing to June.
Is it a service popularity and luck of resources?
Asking for advice how to solve it. will try to downgrade model to previouse version.
1 Reply
endpoint: abm9xoofr5v3rq
as i understand its a problem with serverless infrastructure
removed 4090 from allowed GPU, A40 seems working well.