T‑LLaMA: a Tibetan large language model based on LLaMA2

In this study, we built a corpus containing 2.2 billion Tibetan characters and trained Tibetan LLaMA based on LLaMA2 7B. We achieved state-of-the-art performance in the text classification task using the open-source TNCC dataset, with an accuracy of 79.8%. Additionally, we obtained promising results in text generation and text summarization tasks.

Downloads last month
27
Inference Providers NEW
This model is not currently available via any of the supported third-party Inference Providers, and the model is not deployed on the HF Inference API.