When I run qwen 3.6 27b model on my windows 16gb vram machine, the harness looks like times out when the model tries to write a file. Then it submits the same user query again.
The model run quite slowly (few tps) but works fine on vanilla pi.
Is there a place to configure the timeout?
When I run qwen 3.6 27b model on my windows 16gb vram machine, the harness looks like times out when the model tries to write a file. Then it submits the same user query again.
The model run quite slowly (few tps) but works fine on vanilla pi.
Is there a place to configure the timeout?