Runpod•3mo ago

CUDA Kernel Errors

I am getting the following errors when trying to execute comfy workflows on certain pods.

CUDA error: no kernel image is available for execution on the device
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

CUDA error: no kernel image is available for execution on the device
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

8 Replies

giosuelOP•3mo ago

this happened with the B200 instance as well as with an instance with 2x RTX 5090 it didn't happen with the instance with 2x RTX 4000 ADA with the same workflow

NeuroDragn•3mo ago

oh! plus +1 for me on this issue too, started happening yesterday but doesnt affect all the pods, had to keep switching to find one that worked.

Unknown User•3mo ago

Message Not Public

giosuelOP•3mo ago

I am using the official runpod comfy image, is there a way to create a ticket or something to request this for the official image? It also seems like some comfy components are outdated in that image and also the other problem I am experiencing (https://discord.com/channels/912829806415085598/1391393684192493709) Alternatively, I could use a community image, but there are many different ones and I haven't really tried many of them yet because I'd have to pay for running an instance just to try out the images

Unknown User•3mo ago

Message Not Public

giosuelOP•3mo ago

Is this a community image?

Unknown User•3mo ago

Message Not Public

Madiator2011 (Work)•3mo ago

#Better ComfyUI Slim (5090 supported)

Gaming

Programming

CUDA Kernel Errors

Did you find this page helpful?