[HuggingFace] Hi, i've been working on a new transformer architecture. What datasets should I train it on for basic conversations?

Part of

Inaudible Speech LLM Vulnerabilities