Run quants
#2
by
YuLexuan30
- opened
It is possible to run the quantised models with less than 80GB, I don't think HF even has a single 80GB gpu available for paid users.
Here are the quants, Q6 fits on my 3060
https://huggingface.co/city96/HunyuanVideo-gguf
The A100 is 80GB. Then there are multiple GPU machines that can go over 80GB like:
Nvidia 4xL4
48 vCPU
186 GB RAM
96 GB VRAM