stable-diffusion.cpp
stable-diffusion.cpp copied to clipboard
[Feature Request] Hunyuan Image 3.0 support
https://github.com/Tencent-Hunyuan/HunyuanImage-3.0
Tencent just release an open image model.
Not relevant for us with common consumer hardware.
"System Requirements 🖥️ Operating System: Linux 🎮 GPU: NVIDIA GPU with CUDA support 💾 Disk Space: 170GB for model weights 🧠 GPU Memory: ≥3×80GB (4×80GB recommended for better performance)"
Now this is going to be fun and probably not the best fit for sd.cpp. This is a model that is more fit to be implemented in llama.cpp, I think.
Don't get too deterred by the gpu memory requirements, it has >80B total parameters but only 13B will be active at a time due to the MoE arch.