Go to torrents, get the leak from FB of LLAMA. Scan HN for plenty of entries of GitHub "how to" with said LLAMA. Follow steps there and load the model. See if that's the trigger or this is a hoax.