| license: other | |
| # 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)](https://huggingface.co/922-Narra/cebuano-lm-lora-tests/tree/main/lc2_08212023a) |