Open source models support

[artias13]

Any thoughts on adding support for custom open source models like LLaMa, Alpaca, Alpaca-lora, etc?

[MarkSchmidty]

There's a closed PR from two days ago that added local model support. #29

Not sure what the status is right now. But I don't see anywhere in the current code capable of sending requests anywhere other than the OpenAI API, which is disappointing.

---

Alpaca, Alpaca-lora, etc

On a bit of a tangent, the recently released Vicuna is far superior to Alpaca and other fine tunes. While Alpaca is trained on 60k synthetic question/answer pairs by GPT-3.5 and GPT4All is trained on 400k GPT-3.5 synthetic question/answers, Vicuna is trained on 90k full conversations (multiple question/answers each) between real humans and GPT-4 or GPT-3.5. The dataset for this comes from ShareGPT and can be accessed here: https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/tree/main/HTML_cleaned_raw_dataset

[MarkSchmidty]

This issue might be a duplicate of #25

[spergware]

+1

[Kenshiro-28]

+1

[Pwuts]

Closing as duplicate of #25