Isn't Davinci 003- Amazing?

IMO its a good idea to test out the simplest model that can get the job done, rather than the most complex. My rationale for this is that I want to eventually run these models locally, so I'd want to have a workflow that is cost and resource efficient. It isn't all fine-tuning either, combining 'instruct' methods with few-shot in context learning helps a lot too, too.

Mactronn said:

Datasets and models are available for free on HuggingFace. You can't compare those things.

Anyway, I have a plan to give away access to my fine-tuned models with my custom datasets to 10 BHW members somewhere in March or April.

Click to expand...

I hope locally deployable models I would love to test them out