These are my short notes on how to run ComfyUI on an NVIDIA 5090 GPU. The Blackwell architecture has been released Q4 2024, and is supported by nightly >= 12.8 pytorch builds.

If you get “CUDA error: no kernel image is available for execution on the device” or “NVIDIA GeForce RTX 5090 with CUDA capability sm_120 is not compatible with the current PyTorch installation. The current PyTorch install supports CUDA capabilities sm_50 sm_60 sm_70 sm_75 sm_80 sm_86 sm_90.” errors on a 5090 GPU, and you ruled out an nvidia driver version mismatch between the nvidia-smi reported CUDA version and your installed pytorch version, then the solution is to use a nightly pytorch build that supports CUDA capability sm_120.

Below are steps to install ComfyUI with the correct pytorch version. A Docker container link is available at the bottom of this page:

  1. Clone the ComfyUI repository: git clone https://github.com/comfyanonymous/ComfyUI.git .
  2. Install pytorch and related packages: . venv/bin/activate && pip install --pre torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/nightly/cu129 onnxruntime-gpu numba diffusers
  3. Install ComfyUI dependencies: . venv/bin/activate && pip install -r requirements.txt
  4. Run ComfyUI: . venv/bin/activate && python main.py --listen 0.0.0.0

Here is a Makefile with all of the above:

all:
        python3.10 -m venv venv
        . venv/bin/activate && pip install --pre torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/nightly/cu129 onnxruntime-gpu numba diffusers
        . venv/bin/activate && pip install  opencv-contrib-python==4.8.1.78 opencv-python==4.8.1.78 opencv-python-headless==4.7.0.72
        . venv/bin/activate && pip install -r requirements.txt
run:
        . venv/bin/activate && python main.py --listen 0.0.0.0

Run make all to install dependencies, and make run to start the webui.

Here is the link to my ComfyUI docker container: https://github.com/fgheorghe/comfyui