Sage Attention 3 public release and need to add a node to select attention type on steps
Feature Idea
After multiple months, Sage Attention 3 has been released to the public at https://github.com/thu-ml/SageAttention/tree/main/sageattention3_blackwell.
It employs the hardware-level FP4 cores in Blackwell series GPUs, enabling the attention to proceed x5 times faster than the fastest FlashAttention2 on these GPUs. Comparably to SageAttention2 it's around a x2 times boost.
The caveat is that this precision causes artifacts, especially in the MoE Wan2.2 model (where caching methods such as MagCache and EasyCache also give not great results). Likely it happens because of the MoE timesteps sensitivity, so a Attention Type steps node will probably be an essential addition
Existing Solutions
There is this Pull Request, although it is dated and also needs renaming of the imports https://github.com/comfyanonymous/ComfyUI/pull/9047. Additionally, Kijai has already implemented Sage Attention 3 in his WanWrapper nodes https://github.com/kijai/ComfyUI-WanVideoWrapper/, with the vanilla Sage Attention 2 running on the first and the last steps
Other
No response
Hacked ComfyUI and this is how SageAttention3 landscapes look in Qwen, for reference
Bad luck with characters currently
yeah i want sageattention 3 in comfyui to use it with my rtx 5060 ti xD
just not sure if fp4 is literally gonna work without any calibration or re-training...
How did you make it work on qwen?
I have a PR for this https://github.com/comfyanonymous/ComfyUI/pull/11026