the tokenizer is a bit broken right now, trying to fix it. the model is fully trained and working, just fixing some stuff in the model emulator. if you want to check it out; https://axialm.github.io/chat
