Spaces:
Running
Inference keeps failing because of "ZeroGPU worker error GPU task aborted" and I had my daily ZeroGPU usage consumed
As the title, I paid $9 subscription today and I tried to inference with ZeroGPU, but I always get "ZeroGPU worker error GPU task aborted" error when I inference and it's consuming my 25mins per day quota.
I don't think this is fair since the inference didn't even actually finish and my ZeroGPU usage is getting consumed?
Is there any way I can lower the chance of getting the error?
As the title, I paid $9 subscription today and I tried to inference with ZeroGPU, but I always get "ZeroGPU worker error GPU task aborted" error when I inference and it's consuming my 25mins per day quota.
I don't think this is fair since the inference didn't even actually finish and my ZeroGPU usage is getting consumed?Is there any way I can lower the chance of getting the error?
One of the possible reasons for the above error is that the maximum inference time of ZeroGPU is 5 minutes or you need to increase the duration in @spaces.GPU
@eienmojiki Thanks for the reply, I'm actually running NotaGen-X and the inference of this model is never more than 5 minutes. As far as you know is there any other possible reasons that the error is occurring?
When I tried to reproduce and run it, it simply crashed with an error in the final processing.
https://huggingface.co/spaces/ElectricAlexis/NotaGen
/usr/lib/x86_64-linux-gnu/libjack.so.0
findlib: libnss3.so: cannot open shared object file: No such file or directory
/home/user/app/squashfs-root/AppRun: Using fallback for library 'libnss3.so'
QStandardPaths: XDG_RUNTIME_DIR not set, defaulting to '/tmp/runtime-user'
@John6666
Thanks for trying to reproduce it, I didn't know there's an official Space available, I was using my own version (Modified from the GitHub repo's gradio/ directory and uploaded to the Space).
I think I should try to duplicate the official one and test it, see if the error still exists