I just got myself some phones for research, including ones with Qualcomm, Mediatek, Apple, and Kirin(Huawei) chips. All of them work with MLC except for the one with the Kirin 9000S chip, where it successfully load models, but crashes every time I make a prompt. I’ve also tried previous versions of MLC, and also alternatives like Maid/sherpa. I think it’s a chip issue rather than an environment issue, since the same problem occurs even when I use VMOS.
Have anyone tried running LLMs on the new Huawei phones with Kirin 9000S? If so, did you run into the same issue?
Probably a Vulkan driver quirk. I would report it to Huawei, somehow, if you can. They probably have some business interest in running LLMs on their phones, and would fix it if they knew.