UPDATE!
download my ComfyUI Build https://huggingface.co/datasets/StefanFalkok/ComfyUI_portable_torch_2.10.0_cu130_cp313_sageattention_triton/tree/main , also u need download and install CUDA 13.0 (https://developer.nvidia.com/cuda-13-0-0-download-archive) and VS Code (https://visualstudio.microsoft.com/downloads/)
My TG Channel - https://t.me/StefanFalkokAI
My TG Chat - https://t.me/+y4R5JybDZcFjMjFi
Also big thanks to
for helping with workflows in past!
Hi! I introduce my working workflows with Wan 2.2 generation video for ComfyUI
I have included 4 workflows such as t2v, i2v, v2v, flf2v (firstlastframe) and foley audio
You need to have Wan 2.2 10 steps models (https://huggingface.co/StefanFalkok/Wan_2.2_10steps/tree/main), clip (https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) and vae (wan 2.1 vae)
GGUF Wan 2.2 10 steps models
https://huggingface.co/StefanFalkok/Wan_2.2_10steps_GGUF
Leave comments if you have trouble or you found the problem with workflows
Description
New samplers Res_3m, Res_3m ode, Res_2m, Rs_2m ode
FAQ
Comments (8)
Hi, I am getting this error here. I am guessing it is to do with Torch or Triton, but I do not know what to do. DO I need to update something or is the issue something else. My GPU is a RTX 3080. I have installed the same models as the default json came with.
ERROR:
torch._inductor.exc.InductorError: CompilationError: at 1:0:
def triton_poi_fused__to_copy_add_copy_mul_4(in_out_ptr0, in_ptr0, in_ptr1, in_ptr2, xnumel, XBLOCK : tl.constexpr):
^
ValueError("type fp8e4nv not supported in this architecture. The supported fp8 dtypes are ('fp8e4b15', 'fp8e5')")
Hi. Try install e5m2 models, because e3m4fn models don't support rtx 3000 and older GPU, or install models from Alibaba
falkok21z Seems to fix, but I think my GPU is just bad, I checked the logs as it processed and it is sending this (and lots of other stuff):
:2691] [6/0_1] No valid triton configs. OutOfMemoryError: out of resource: triton_mm Required: 131072 Hardware limit:101376 Reducing block sizes or num_stages may help..
And a bunch of other text and then it got stuck. Thanks though, probably need an upgrade gpu.
TheBagOfPopcorn you can try using without torch , but yep - it's right answer regarding the upgrade of GPU, at least rtx 4000 series. I have talked many people who has rtx 3000 GPU and I listened many complains in generating wan 2.2 videos
falkok21z Yep, probably wont happen for a while, not that economically fortunate. But honestly my Wan2.2 T2V and I2V have been decent so far. Thanks for the help either way!
falkok21z Also I wanted to ask about VACE, because that model works good for V2V and I am able to create vids with that which properly use controlNet. But the quality is honestly much worse than 2.2. Is that something that everyone is suffering or I am downloading a wrong model perhaps? I am using Q8 gguf.
TheBagOfPopcorn I'm also waiting WAN 2.2 VACE model to update my workflows, but 2.2 VACE still isn't existing at the moment. These VACE models you have installed are probably have worth quality , because they are unofficial models
falkok21z Makes sense yea.