THIS PROJECT IS NO LONGER IN USE. For a quicker and local version, check out my other project: https://scratch.mit.edu/projects/1102581816/ Yep. Llama2 in scratch. Llama 2 is similar to ChatGpt model 2. Because I am not that great at Python, this is limited to one person using it at a time. ------------------------DISCLAIMER---------------------------------- I can see everything you type into this. It is supposed to keep responses short, so it can fit the response in the cloud variables. Will not work for New Scratchers, nor non signed in users. I would love to hear feedback/questions/suggestions down below. :) Sorry if it runs into errors. If the first response is blank, say the same thing. It should work after that. It will automatically check for inactivity, so people can use it. If this gets enough support, I will upgrade to Llama 3.8 (Similar to ChatGpt 3.5)
Thanks to TimMcCool for the ScratchAttach python package that makes this possible. I used llama 2 from Huggingface.co. Llama 2 is like ChatGpt 2. I also used Ollama from ollama.com to run the Gpt. Llama 2: https://huggingface.co/meta-llama/Llama-2-7b-chat-hf Ollama: Ollama.com ScratchAttach: https://github.com/TimMcCool/scratchattach/wiki This is all hosted on a small computer at my house. It is not very powerful, and my internet is kinda slow, so responses can be slow. V1.2: Added Longer output support V1.1: Simple Bug fixes and faster inactivity detection. Added the chatgpt logo instead of black screen. V1: Release! First working version.
