[Badcase]: repetition after the conversation and send endless emoji (Ollama + OpenwebUI)

[valleysprings]

Model Series

Qwen2.5

What are the models used?

Qwen2.5-72B

What is the scenario where the problem happened?

Qwen2.5-72B gets repetitive and send endless emoji after the useful conversation ends

Is this badcase known and can it be solved using avaiable techniques?

• [X] I have followed the GitHub README.
• [X] I have checked the Qwen documentation and cannot find a solution there.
• [X] I have checked the documentation of the related framework and cannot find useful information.
• [X] I have searched the issues and there is not a similar one.

Information about environment

OS: Ubuntu 22.04 Python: Python 3.12 GPUs: 2 x NVIDIA A100 NVIDIA driver: 560 CUDA compiler: 12.6 Ollama 0.3.11

Description

Steps to reproduce

This happens to Qwen2.5-72B The badcase can be reproduced with the following steps:

1. no system prompt
2. default hyperparameter

The following example input & output can be used:

1. 条件熵 相对熵 互信息 kl 散度 定义 和 联系
2. 怎么证明互信息是从条件熵推过来的 计算一下

Expected results

The results are expected to be no repetition and end the conversation.

Anything else helpful for investigation

I find that this problem also happens to ...