Oudated Playbook: https://build.nvidia.com/spark/comfyui/video-gen-workflow

Accelerated Computing DGX Spark / GB10 User Forum DGX Spark / GB10
1

Hi All,

Purchased my Spark yesterday. New to forums and AI so forgive me if this is the wrong place or manner to ask for help.

I’m trying to follow the playbook listed here: https://build.nvidia.com/spark/comfyui/video-gen-workflow. It seems like this playbook is out of date for the current version of the Spark DGX, and I’m having trouble running the Playbook to completion. For instance, Step 3 instructs:

git clone GitHub - NVIDIA/dgx-spark-playbooks: Collection of step-by-step playbooks for setting up AI/ML workloads on NVIDIA DGX Spark devices with Blackwell architecture. · GitHub
cd dgx-spark-playbooks/nvidia/playbook-comfyui
docker build -t comfyui -f assets/Dockerfile .

But, when I clone the repo, the listed directory is wrong:

Expected: dgx-spark-playbooks/nvidia/playbook-comfyui
Actual: dgx-spark-playbooks/nvidia/comfy-ui

Additionally, the docker command seems to expect a dockerfile within the assets directory:

docker build -t comfyui -f assets/Dockerfile .

But, I don’t see a dockerfile file or directory within the assets dir, only some setup scripts:

~/dgx-spark-playbooks/nvidia/comfy-ui$ ls -lhat ./assets/
total 16K
drwxrwxr-x 2 [USER GROUP] docker 4.0K Aug 24 12:07 .
-rwxrwxr-x 1 [USER GROUP] docker 1002 Aug 24 12:07 setup.sh
drwxrwxr-x 3 [USER GROUP] docker 4.0K Aug 24 12:07 ..
-rwxrwxr-x 1 [USER GROUP] docker   99 Aug 24 12:07 launch.sh

What am I doing incorrectly, if anything? Does anyone else have this problem? Has anyone successfully ran the current playbook to completion without issues?

2

Thanks for raising this, definitely seems outdated. I will reply with instructions to get you unblocked and work with the team to update the playbook

3

Playbook has been updated with better instructions and the assets directory is back for you to build the Docker image

4

That Dockerfile uses nvcr.io/nvidia/pytorch:26.02-py3 but if I try to run that container, I see this error:

ERROR: This container was built for NVIDIA Driver Release 590.48 or later, but
       version 580.173.02 was detected and compatibility mode is UNAVAILABLE.

I’ve installed all updates. Is there something else I need to do?

(I’m not currently using the Dockerfile, I was just trying to figure out what the latest pytorch image that worked was and noticed that the one here didn’t seem to work)

5

This could be an issue with the NVIDIA container toolkit or with the injection of the GPU into the container. It shouldn’t be a problem with the container itself. Make sure you are properly using --gpus '"device=0"' in your docker run command