ComfyUI-Lightning

The codes are modified from SageAttention, ComfyUI-TeaCache, comfyui-flux-accelerator and Comfy-WaveSpeed.

I use follow tips to accelerate FLUX inference speed:

1. Add TeaCache or FBCache;
2. Skip some unnessasery blocks;
3. Compile and quantize model;
4. Use fast CuDNN attention kernels;
5. Use SageAttention.

More details see above given repositories.

Usage