Support of Beta API on vLLM

#21
by taoqiuyang - opened

We are running DeepSeek-v2.5-236B locally with using vLLM on 8*80GB GPUs as suggested.
It seems the DeepSeek Beta API (with FIM and other new features) are not supported, only v1 API is available.
Is there a way to use DeepSeek Beta API on local GPUs with vLLM or other inferencing libraries?

Sign up or log in to comment