lllyasviel
5070 Ti not supported? ForgeUI fails with TypeError and CUDA kernel errors
Hi ForgeUI team,
I recently upgraded my GPU from a Zotac RTX 3060 to a Zotac RTX 5070 Ti Amp Extreme (Blackwell architecture) and have been trying to get ForgeUI working on it for the past couple of days with no success.
Here are the issues I’m facing:
I followed the instructions to install Python 3.10.x, created a virtual environment, installed Forge UI, and installed the correct nightly version of PyTorch with CUDA 12.8 support.
I also made sure CUDA 12.8 is installed on my system.
I’ve reinstalled Python, ForgeUI, Torch/Torchvision, InsightFace, etc. multiple times.
When launching ForgeUI, I get this error consistently:
TypeError: 'NoneType' object is not iterable
CUDA error: no kernel image is available for execution on the device
I’m reaching out here after exhausting Reddit, ChatGPT, and every workaround I could find. Can someone please confirm if 50-series (Blackwell) GPUs like the 5070 Ti are currently supported? If not, is there a patch or dev branch being worked on?
Any help or even acknowledgment would mean a lot — I’m happy to test things or share logs if needed.
System Info:
-
GPU: Zotac RTX 5070 Ti Amp Extreme 16GB
-
CUDA: 12.8
-
Python: 3.10.6
-
PyTorch: 2.5.1+cu128 (nightly)
-
OS: Windows 10
Thanks in advance!
PyTorch: 2.5.1+cu128 (nightly)
I don't think there is a cu128 version of PyTorch 2.5.1. The version with 12.8 support is on the preview build that comes after 2.6.0. Was this the pip command you ran?
pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu128
Just in case this answer provides the step by step: https://github.com/lllyasviel/stable-diffusion-webui-forge/issues/2812#issuecomment-2817162925. Just be sure you're installing to the venv.
there is now, activate your venv, then do
pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128
