Is the STQ runtime kernel for the 1.25-bit model open-sourced?

#3
by baicai1145 - opened

I can see the Sherry paper/code and the Hy-MT 1.25-bit / GGUF releases.

I want to confirm one point:
is the actual STQ runtime kernel used for 1.25-bit deployment open-sourced?

From what I can tell:

  • the Sherry repo contains training/reference code,
  • the Android demo APK appears to use GGUF + llama.cpp/ggml,
  • but I could not find the open-source STQ mobile/runtime kernel itself.

Could you clarify whether:

  1. the STQ kernel is already public,
  2. it will be open-sourced later,
  3. or the current community-available deployment path is GGUF only?

Thanks.

AngelSlim org

STQ is not public yet, it will be released soon.

Sign up or log in to comment