Can we use 1.5B model like Qwen 1.5B in this competition? Does the parameter size of non-llama must STRICTLY under 1.5B? Or we can use 1.5B models as well.
