Debjit Biswas

Congratulation! You are a problem solver.

I think 2GB for 350M model can run, Here is the function that defines it. ``` https://github.com/fauxpilot/fauxpilot/blob/770f56a6588fbcf1a366c3b9ec9efc4701f52943/setup.sh#L59 ``` ``` echo "Models available:" echo "[1] codegen-350M-mono (2GB total VRAM required; Python-only)"...

The server was deleted as I was running it as stand-alone using SQLite and my antivirus sees it as a threat and deleted it. I was debugging using dd so...

Awesome! You are a hero.

Congratulations! You have solved it.

Congratulations! You have done it.

When I was working with other compatible OpenAI APIs, I also discovered this issue. Should have been marked.

So we still do not have any option to pass a custom URL for self-hosting LLM? Or just have to pass as an HTTP request?

An official prebuilt docker image will be way more helpful. I will try this later in the evening.

The issue is still present. Same issue.