unsloth

/

Qwen2.5-VL-7B-Instruct-GGUF

like 54

Follow

Unsloth AI 8.97k

Hi, I’m running Qwen2.5-VL-7B-GGUF with llama-server.
In the logs I see:

--- vision hparams ---
load_hparams: image_size:         1024
load_hparams: patch_size:         14

I would like to increase the image_size to 2048.
Is there a way to override this at runtime (like a flag --image-size).