WangchanLION is a joint effort between VISTEC and AI Singapore to develop a Thai-specific collection of Large Language Models (LLMs), pre-trained for Southeast Asian (SEA) languages, and instruct-tuned specifically for the Thai language. WangchanLION-v3 is a multilingual model that has been continual pre-training with around 47.4 billion Thai samples from web and non-web data — Developed by: Products Pillar, AI Singapore, and VISTEC — Funded by: Singapore NRF, PTT Public Company Limited, SCB Public Company Limited, and SCBX Public Company Limited — Model type: Decoder — Languages: English, Thai — License: Llama 3.1 Community License We performed continual pre-training in Thai on Llama 3.1 8B CPT SEA-LIONv3 in For tokenization, the model employs the default tokenizer used in Llama3.1-8B. The model has a context length of 128k. NOTE This model has not been trained to use a system prompt or to use tool calling. Therefore, it can only use with the supervised fine-tuning (SFT). It is important for users to be aware that our model exhibits certain limitations that warrant consideration. Like many LLMs,…
Модальности:
Генерация текста
Области применения:
Диалог / чат
Задача: Генерация текста
Автор: aisingapore
Теги: llama, conversational, en, th, text-generation-inference, endpoints_compatible
Лайков: 4 | Загрузок: 22
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
