Can it work with Llama 3 / other 7b models?

[pjbaron]

Could this process be used on Llama 3 8b? I find that "Llama 3 is an auto-regressive language model that uses an optimized transformer architecture." I have a task to update the knowledge in Llama 3 for the Phaser game engine. It currently gives mixed responses using both Phaser 2 and Phaser 3 code examples. We'd like it to either learn the difference or forget everything it 'knows' about Phaser 2 (and 3 if necessary - I have an effective RAG to provide Phaser 3 knowledge in this case). (Llama 3 is just a 'for instance', we're not locked into any single model at the moment. I'm about to start trying a variety of them to see if there's something more suited to this task).