i thought a possible spam thread could be talking to a language model. i thought i’d try the newish yarn model with 128k context length maybe i could run it in colab or i dunno.
possible model: https://huggingface.co/NousResearch/Yarn-Llama-2-13b-128k smaller quantized variant of model: https://huggingface.co/TheBloke/Yarn-Llama-2-13B-128K-GPTQ the quantized one is under 8GB. there are also smaller variants.. people are likely hosting it on free api servers somewhere. i’m not presently looking at that.
