Warning: There are some problems with the tokenizer of this model, which will be corrected in the next version of the model (chat-informer-1b). We are honored to introduce a lightweight Chinese-English conversation assistant designed to reduce the cost of inference. It is trained from scratch, based on the LLAMA2 architecture, with 150 million parameters and a completely new vocabulary. The training process consists of two parts: (1) NTP task. (2) Instruction tuning. The model improves data quality for pre-training and instruction tuning.
Модальности:
Генерация текста
Области применения:
Диалог / чат
Задача: Генерация текста
Автор: liftkkkk
Теги: llama, en, zh, text-generation-inference, endpoints_compatible
Лайков: 2 | Загрузок: 10
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
