昇腾lmdeploy使用 lmdeploy APIClient 接口时，推理结果被截断

[winni0]

Checklist

• 1. I have searched related issues but cannot get the expected help.
• 2. The bug has not been fixed in the latest version.
• 3. Please note that if the bug-related issue you submitted lacks corresponding environment info and a minimal reproducible demo, it will be challenging for us to reproduce and resolve the issue, reducing the likelihood of receiving feedback.

Describe the bug

昇腾lmdeploy使用lmdeploy serve api_server
/LLaMA-Factory-main/model/Qwen2.5-7B-Instruct
--backend pytorch
--server-port 8000
--device ascend
--session-len 8192启动服务，使用 lmdeploy APIClient 接口接收结果时，推理结果被截断。

Reproduction

API服务启动命令如下：lmdeploy serve api_server
/LLaMA-Factory-main/model/Qwen2.5-7B-Instruct
--backend pytorch
--server-port 8000
--device ascend
--session-len 8192，lmdeploy APIClient 接口代码如下：

Environment

TorchVision: 0.18.1
LMDeploy: 0.6.4+191a7dd
transformers: 4.47.1
gradio: Not Found
fastapi: 0.115.6
pydantic: 2.10.4
triton: Not Found

Error traceback

没有报错

[jinminxi104]

please add max_tokens into completions_v1. (please also note that you are using graph mode, which cause a compile phase at first run.)

[winni0]

你这个回答结果也是被截断的呀

[jinminxi104]

你这个回答结果也是被截断的呀

我也设定了截断的值。你可以设大，比如设定2k

[MiningIrving]

我也设定了截断的值。你可以设大，比如设定2k

请问，现在支持qwen2.5:14B及以上参数的模型吗

[jinminxi104]

我也设定了截断的值。你可以设大，比如设定2k

请问，现在支持qwen2.5:14B及以上参数的模型吗

支持，请设定合适的tp值。