ollama - 💡(How to fix) Fix Running qwen3.6:27b-bf16 on an AMD Ryzen AI Max leads to gibberish [1 participants]

TL;DR

The issue might be resolved by adjusting the VRAM allocation or checking the compatibility of the "qwen3.6:27b-bf16" model with the Ollama version 0.21.2.

Guidance

• Verify that the "qwen3.6:27b-bf16" model is compatible with the Ollama version 0.21.2, as other quantized models run without issues.
• Check if reducing the dedicated VRAM from 96 GB to a lower value improves the performance or resolves the gibberish output.
• Investigate if there are any known issues or limitations with the GMKTec EVO X2 (AMD Ryzen AI Max) hardware configuration that could be affecting the model's execution.
• Review the Ollama documentation for any specific requirements or recommendations for running the "qwen3.6:27b-bf16" model.

Notes

The provided information lacks specific error messages or logs, making it challenging to pinpoint the exact cause. Further investigation into the model's compatibility and hardware configuration is necessary.

Recommendation

Apply workaround: Adjust the VRAM allocation or model configuration to see if it resolves the issue, as the problem seems to be specific to the "qwen3.6:27b-bf16" model.