Cebu-llama-v0.1

• Experimental Cebuano fine-tune: safe or accurate outputs not guaranteed (not for production use)!
• Observing fine-tune effects on dataset of roughly 10k lines of Cebuano (roughly formatted/"raw" chat)
• Trained on llama 7b chat for 1 epoch
• May still generate English, Tagalog, Taglish, or gibberish.
• QLoras (hf and GGML)