ilab model chat times out while waiting for vLLM to start up. Bumping from the default 120 to something like 320 is necessary. The usual retry count varies around 280 attempts.

eg.

serve.vllm.max_startup_attempts: 320