Another experimental model, using mostly sythetic data generated by airoboros This is essentially a minor «fix» branch of airoboros-l2-13b-2.2 with a updates, primarily: — re-generated writing responses — longer contextual blocks — removal of «rp» data — (less aggressive) de-censoring — more fine-tuning epochs This is a fairly general purpose model, but focuses heavily on instruction following, rather than casual chat/roleplay. Huge thank you to the folks over at a16z for sponsoring the costs associated with building models and associated tools! The default system prompt («A chat.») was used for most of the prompts, however it also included a wide sampling of responses with other prompts, particularly in «stylized_response», «rp», «gtkm», etc. And chat scenario that wouldn’t require USER/ASSISTANT (but should use stopping criteria to prevent the model from speaking on your behalf). I strongly suggest adding stopping criteria/early inference stopping on «USER:», and/or whatever names you specify in the system prompt. https://wandb.ai/jondurbin/airoboros-l2-13b-2.2.1/runs/zbz8mgaz?workspace=user-jondurbin The prompts shown here are are just the text that would be included after…
Модальности:
Генерация текста
Задача: Генерация текста
Автор: jondurbin
Теги: llama, text-generation-inference, endpoints_compatible
Лайков: 3 | Загрузок: 61
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
