[Badcase]: 触发模型不断生成121121121或则其它乱码，这个正常吗

[Liubo111111111]

Model Series

Qwen2.5

What are the models used?

Qwen3-235B-A22B

What is the scenario where the problem happened?

Qwen平台

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

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

Information about environment

触发模型不断生成121121121或则其它乱码，这个是因为超出模型生成上下文导致的吗

Description

https://chat.qwen.ai/s/759820b1-c984-4c7f-9af7-ffa1b4d3045c?fev=0.0.85
生成badcase聊天

[denglj]

我也遇到了相同的问题。qwen3 14b。

[science2011]

Qwen3-30B-A3B 也遇到了相同的问题，部署框架MindIE，硬件800A2

[jklj077]

我也遇到了相同的问题。qwen3 14b。

Tibetic languages are not officially supported. The models may know about the languages from its training data but the performance is not validated.

It is recommended to try a larger model for this usecase.

For Ollama, please consider using a higher quant and set the sampling parameters accordingly.

[jklj077]

Qwen3-30B-A3B 也遇到了相同的问题，部署框架MindIE，硬件800A2

Without steps to reproduce, it is hard to know whether it is a model issue or a framework issue. It is recommended to seek help from MindSpore on this matter first.

[jklj077]

触发模型不断生成121121121或则其它乱码，这个是因为超出模型生成上下文导致的吗

We are looking into this. Since there are other logics at Qwen Chat (context management, etc.), it would need more time, but the context length should be an important factor here.

[StarMagicAI]

me too

[jklj077]

me too

what's the contexts/prompts?