A report about the bugs on Qwen2.5VL 7b

Note: I read some similar issue and tried to provide more detailed imformation to solve the problem on this one.

I use NVIDIA RTX 3070 Laptop with Game Ready Drive 572.83 on Windows11 24h2. And I use the latest stable version of LM Studio(0.3.14 build5, v1.26.0 runtime).

I got the following error when loading Mungert/Qwen2.5-VL-7B-Instruct-GGUF/Qwen2.5-VL-7B-Instruct-iq4_xs.gguf

🥲 加载模型失败

Error loading model.

(Exit code: 18446744072635810000). Unknown error. Try a different model and/or config.

I tried different runtimes and got the same exit code, but it's different than ouher user's.

Then I tried Misaka27260/Qwen2.5-VL-7B-Instruct-abliterated-GGUF, It works properly using different runtimes. I only find the two versions of Qwen2.5-VL-7B with vision support, the other users that got similar error may all used Mungert/Qwen2.5-VL-7B-Instruct-GGUF/Qwen2.5-VL-7B-Instruct. So I think it may be an issue with quantalization. It's adviced to try different configs of quantalization and provide an official version that works properly.

Start by reproducing the failure with Mungert/Qwen2.5-VL-7B-Instruct-GGUF/Qwen2.5-VL-7B-Instruct-iq4_xs.gguf on LM Studio 0.3.14 with the listed Windows 11 and RTX 3070 setup. Compare the failing model across runtimes with the working Misaka27260 model, then identify whether the quantization or model compatibility causes exit code 18446744072635810000 and document a confirmed workaround or fix.