Spaces:

Alpha-VLLM

/

Lumina-Next-T2I

like 81

Running on Zero

Hi @PommesPeter , I think the GPU task aborted error is raised when the function decorated with @spaces.GPU takes longer to run than the specified time limit, which is 60 seconds by default. Can you try and increase the limit using duration parameter, for example, like @spaces.GPU(duration=120). I've restarted the demo as well, hoping to get a clearer message in the logs this time.

Hi @ysharma ,

thank you for your reply. i got it. i have solved this problem. The error was caused because our previous logic required the model to remain on the GPU continuously. However, space.GPU appears to call the GPU dynamically, so we couldn't keep the model loaded on the GPU at all times. We've now modified the logic to load the model onto the GPU only when inference is called.

Now, it works. you can try our model.

https://huggingface.co/spaces/Alpha-VLLM/Lumina-Next-T2I

This is great news! Thanks for letting me know, and congratulations on a great release.