ollama - 💡(How to fix) Fix Running qwen3.6:27b-q8_0 produces also gibberish on an AMD Ryzen AI Max+ [1 participants]

TL;DR

• The issue might be related to compatibility or configuration problems between qwen3.6:27b-q8_0 and ollama 0.22.0, considering the gibberish output of random ASCII characters.

Guidance

• Verify that the versions of qwen and ollama are compatible, as the issue seems to persist across different versions of qwen.
• Check the system's GPU and CPU utilization to ensure that the hardware is not overwhelmed, given the high VRAM and the specific AMD Ryzen AI Max+ setup.
• Consider testing with a different version of ollama or qwen to isolate if the issue is version-specific.
• Review any available documentation for qwen and ollama for specific requirements or known issues related to AMD hardware or Windows operating system.

Notes

• The lack of specific log output or error messages makes it challenging to pinpoint the exact cause, suggesting that further debugging or logging might be necessary.
• The issue's description does not provide enough detail to recommend a specific code change or configuration adjustment.

Recommendation

• Apply workaround: Try adjusting the version of qwen or ollama to a previously known stable combination, if available, to see if the issue resolves, as the current combination seems to produce inconsistent output.