用deepseek-coder-instruct聊天模板注册的DeepSeek-Coder-V2-Lite-Instruct模型给出的代码在聊天界面会出现&quot，代表“，但是如果用post得到的代码就不会

[MOON1234567890A]

System Info / 系統信息

cuda版本：
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2021 NVIDIA Corporation
Built on Thu_Nov_18_09:45:30_PST_2021
Cuda compilation tools, release 11.5, V11.5.119
Build cuda_11.5.r11.5/compiler.30672275_0

transformers版本：
Name: transformers
Version: 4.45.2
Summary: State-of-the-art Machine Learning for JAX, PyTorch and TensorFlow
Home-page: https://github.com/huggingface/transformers
Author: The Hugging Face team (past and future) with the help of all our contributors (https://github.com/huggingface/transformers/graphs/contributors)
Author-email: [email protected]
License: Apache 2.0 License
Location: /anaconda3/envs/xinf/lib/python3.10/site-packages
Requires: filelock, huggingface-hub, numpy, packaging, pyyaml, regex, requests, safetensors, tokenizers, tqdm
Required-by: auto_gptq, autoawq, chattts, compressed-tensors, FlagEmbedding, nemo_text_processing, optimum, peft, sentence-transformers, transformers-stream-generator, vllm

python版本：
3.10.15
操作系统：
linux

Running Xinference with Docker? / 是否使用 Docker 运行 Xinfernece？

• docker / docker
• pip install / 通过 pip install 安装
• installation from source / 从源码安装

Version info / 版本信息

1.2.1

The command used to start Xinference / 用以启动 xinference 的命令

xinference-local --host 0.0.0.0 --port 9997

Reproduction / 复现过程

Expected behavior / 期待表现

[qinxuye]

之前 deepseek-r1-distill 模型会输出 <think> 标签，在 gradio 里会被吞掉，因此做了 html.escape。

inference/xinference/core/chat_interface.py

Lines 141 to 145 in 5abfc84

	# some model like deepseek-r1-distill-qwen
	# will generate <think>...</think> ...
	# in gradio, no output will be rendered,
	# thus escape html tags in advance
	response_content += html.escape(delta["content"])

这个需要看下怎么解。