Right now, running the LoRA unit test on GPU gives: ``` Model OPTForCausalLM does not support LoRA, but LoRA is enabled ``` Need to integrate with a LoRA for a smaller but different (ideally smaller) model