They do maintain the Transformers library which is pretty much the core library for how you interact with LLM models in the open source world. So while they weren't using a model they've trained, they were a part of making just about all of the open models (maybe excluding OpenAI and Google's, I wouldn't be surprised if they have their own frameworks that predate the Transformers library).
reply