LiteLLM seems to kill the request after 60s timeout - what to do?

[eleaner]

Hi,
I am running local llm on CPU only (super slow I know but it does not matter for my usecase)
when I run inference directly it just waits (300s for '2+2=?' inference)
but when I proxy the request through litellm - it get's killed after 60s and I believe and it tries again.
I have

router_settings:
  timeout: 1800

litellm_params:
  timeout: 1800

what am I missing?

[eleaner]

I also added

litellm_settings:
  request_timeout: 1800
  stream_timeout: 1800

but no improvement