Twinkle AI

Twinkle AI 是一個在 2024 年誕生的繁中語言模型研究社群，起步於 Llama 開源模型，逐步打造出專屬於繁體中文的落地技術。社群成員來自各行各業，因熱愛模型訓練而聚，從蒐集繁中語料到模型訓練。社群旨在用實際行動推廣大語言模型的訓練知識，並希望透過開放與共享，推動臺灣在生成式 AI 領域的發展。 Twinkle AI is a Traditional Chinese language model research community founded in 2024. It began with the open-source Llama model and gradually developed practical technologies tailored for Traditional Chinese. The community is composed of members from diverse backgrounds who share a passion for model training. From collecting Traditional Chinese corpora to training models, Twinkle AI has already open-sourced Twinkle Eval (an efficient evaluation framework), Traditional Chinese training datasets, and a Traditional Chinese large language model. The community is dedicated to promoting knowledge of large language model training through practical action and aims to advance Taiwan’s development in the field of generative AI through openness and collaboration. 🤗 https://huggingface.co/twinkle-ai 😺 https://github.com/ai-twinkle