Upload fine-tuned Gemma-7B CEFR model with exact and within-1 confusion matrices

initial commit