Initial upload: Llama 3.1-8B LoRA r=16 adapter (3-epoch SFT on rail defect reports)

Add model card

initial commit