Method __encode_base64

fastchat/serve/model_worker.py:178–182

(self, embeddings: torch.Tensor)

        return sum_embeddings, token_num

    def __encode_base64(self, embeddings: torch.Tensor) -> List[str]:
        embeddings = embeddings.cpu()
        return [
            base64.b64encode(e.numpy().tobytes()).decode("utf-8") for e in embeddings
        ]

    @torch.inference_mode()
    def get_embeddings(self, params):