Exp fine-tune the Llama 3.3-8B model for raw, uncleaned dataset include code, math, roleplay with random config LoRA
