TTI(Text 2 Image) inference using stable-diffusion.cpp doesn't work properly on Android phone

[jeffzhou2000]

The Project stable-diffusion.cpp is excellent and thanks for your amazing stable-diffusion.cpp. I want to utilize the stable-diffusion.cpp as TTI(Text 2 Image) engine in my learning project.

TTI(Text 2 Image) inference using stable-diffusion.cpp doesn't work properly on Android phone currently https://github.com/zhouwg/kantv/blob/master/core/ggml/stablediffusioncpp/model.cpp#L773

https://github.com/zhouwg/kantv/blob/master/core/ggml/stablediffusioncpp/stable-diffusion.cpp#L189

I found that GGML's internal / GGML API changes quickly/frequently. could you help to merge upstream GGML source code to stable-diffusion.cpp?

thanks for your time and thanks so much.

[GainLee]

The repo uses ggml @ 57869ad, you should not use the latest version of ggml

[jeffzhou2000]

The repo uses ggml @ 57869ad, you should not use the latest version of ggml

thanks for you comment and thanks for your time.

I see. but the whisper.cpp and llama.cpp couldn't works with ggml @ 57869ad . the whisper.cpp and llama.cpp only works with the latest ggml cause of internal codes of llama.cpp and API of ggml library changs quickly and frequently.

btw, is there any plan to implement StableDiffustion using NCNN?