Continued pre-training of Llama 3.1 70B with expanded tokenizer on Irish human text collection using LoRA