LLM Output is truncated

[fka-at]

Describe the bug

When running LLMs (Deepseek R1) from inside the AI Playground, the output seems to stop / get truncated after approximately 25 sec..
This is noticable when running Deepseek R1 which ouputs a lot of inner monologue. The output suddenly stops.
(According to the GPU-Utilization in Taskmanager, it appears, the model might keep running in the background, but its ouput is not displayed.)

To Reproduce

Steps to reproduce the behavior:

1. Go to the Answer tab
2. Load Deepseek-R1-Distill-Qwen-7B
3. Ask it literally anything that makes it think more than a few seconds
4. wait and observe

Expected behavior

A complete Output

Environment (please complete the following information):

• OS: Windows 11 24H2
• GPU: Intel Arc A770 16GB
• CPU: Ryzen 5800X
• RAM: 32GB
• Version: 2.0.0 alpha
• Backend: IPEX-LLM