pruned ??
would be great if be available Pruned version too :]
Hi @davoodkharmanzar ,
May I ask what do you mean by pruned version?
If you're talking about quantised version, you can try merging the lora into base model first and then apply quantization algorithms on it.
The degree that the quality degrades with quantization should be similar as the base model.
Since the current version is an early preview where the texture details remain limited, we're focusing on this now.
Literally we don't request higher requirements on compute resources versus base model, maybe even lower since we process one chunk by one dit forward.
Everyone can give it a try but currently we're still trying to improve texture details as first priority.
Thanks for your attention!
During this time, I might get ComfyUI node available, which you can expect soon.
Best,
Yanzuo
We’ve released ComfyUI nodes for RAVEN streaming generation, enabling 192-frame 1376×768 T2VA generation within a 24 GiB VRAM envelope. This path still places a substantial demand on system RAM, and further memory optimizations are on the way.
@oliveryanzuolu
This is a really interesting project. I hope that you can eventually make it work with the INT8 pruned version too.
May I ask what do you mean by pruned version?
Information on the pruned version: https://blog.comfy.org/p/minimax-h3-day-0-support-in-comfyui