I'm excited to see what the OpenAssistant crowd does with these models, they seem to have gathered the dataset to finetune them. Lots of people use these model

I'm excited to see what the OpenAssistant crowd does with these models, they seem to have gathered the dataset to finetune them. Lots of people use these models as talk therapy. We really need 1) standalone options, 2) reproducible weights with crowd sourced datasets to reduce biases (or at least know who you're talking to).