start a supported embeddings model server with `vllm serve`, e.g. # The OpenAI client does not support the bytes encoding_format. # The OpenAI client does not support the embed_dtype and endianness ...