get rid of upstream code in __main__.py

[dtrifiro]

Once vllm-project/vllm#5090 (review) is merged, we should be able to use the same engine to spin up both http and grpc servers without duplicating upstream code

[dtrifiro]

@prashantgupta24 the above has been merged, we can move on and remove the upstream code from __main__.py

[prashantgupta24]

Uhh I'm still reading up on asyncio, but the problem that I'm facing right now is this:

https://github.com/vllm-project/vllm/blob/main/vllm/entrypoints/openai/api_server.py#L236-L239
Based on these lines, the run_server function uses asyncio.get_running_loop() to get the event_loop coming from our __main__.py and tries to do event_loop.run_until_complete which is incorrect.

With my limited understanding, this kind of makes me think we won't be able to use asyncio's event loops to start the http_server unless we make changes to upstream?

[dtrifiro]

Yeah, that's correct 😔