fix: correct training config (batch 2x8=16, LR 2e-5, epochs 5)

Added Dataset

Upload folder using huggingface_hub

initial commit