codercowmoo

bait and switch

Depending on the finetuning tool you're using, you can just start the training run, and then it shows you how long it'll take. Like give it 5 mins to stabilise, then see the estimated duration.

Axolotl is a good finetuning tool if you need one.

I really wish people developing local applications would allow users to specify an API endpoint. Most applications use an OpenAI compatible API, and if they don't the browser's implementation of local model inference can be used.

Distil-whisper is incredibly fast. Realtime on a 3060 Ti, and I used it to transcribe an 11 hour audiobook in 9 minutes.

Current SOTA open source is I believe SUPIR (Example - https://replicate.com/p/okgiybdbnlcpu23suvqq6lufze), but it needs a lot of VRAM, or you can run it through replicate, or here's the repo (https://github.com/Fanghua-Yu/SUPIR)

Anyone see the stonks image hidden quite well behind the first table?

I could only see it because of my dark mode extension, otherwise I guarantee I wouldn't have caught it.

Assumedly by scraping it into a raw text corpus format, and feeding it to a model like LLaMA-2.

For those of you who haven't read to the end of the README, its for procrastination. The idea is that if you, as a part of your work, read a lot of papers, but want to procrastinate secretly, you can convert a book or website you'd like to read to a paper, and thus it seems as if you're doing your work but instead you're enjoying your book.

Can you provide some of these sources?