Open acs opened 1 year ago
StableDiffusion (this is not text, but images) BLIP-2 https://github.com/nomic-ai/gpt4all #9 https://laion.ai/blog/open-flamingo/ (vision-language models) https://github.com/hpcaitech/ColossalAI #10 (the best one to learn the architecture behind LLM) https://github.com/microsoft/DeBERTa
An alternative for ChatGPT with the goal of having your own service:
https://github.com/nsarrazin/serge https://github.com/aylabs/chatgpt/issues/8
Let's explore it a bit! It is not open source, but you can download the models and build your own service. It is based in the model: https://huggingface.co/decapoda-research/llama-7b-hf-int4#llama-model-card (also in Alpaca?)
LLama (Large Language Model) comes from Facebook and it was leaked to internet? Alpaca is from Standford university. Both are at GPT3 level.
Some quick intro about them: https://youtu.be/YBWM5Uz-wd8
Alpaca is not available any more because problems with hallucinations.
The goal of Alpaca and it seems that also LLama is to have LLM in low computer hardware. Interesting reading: https://news.ycombinator.com/item?id=35136624
Transferring knowledge between models is easier than expected: https://www.reddit.com/r/singularity/comments/11tmkqi/alpacapromptchaining_considerations_on_the_latest/
AI models training each other
During this trip, nice dissemination article in spanish: https://www.youtube.com/watch?v=FdZ8LKiJBhQ
Talking with jgb he has a presentation about this topic, specially focused in StableDiffusion:
https://jgbarah.github.io/presentations/stable-diffusion/slides-2022-12-01.pdf
The video for the talk: https://m.youtube.com/watch?v=NOGB2BrDG88