A user on the Hugging Face forums expresses frustration that the platform's "Zero GPU" limit applies even when image or video generation tasks fail due to errors or bugged generators. The poster argues that this approach wastes valuable computing resources and results in a 24-hour wait without producing any output.
- Failed generation attempts are counted toward the strict usage limits alongside successful ones.
- The user describes the current system as a "rough sledgehammer method" that lacks distinction between success and failure.
- The complaint highlights the inability to distinguish bugged processes from valid requests.
The post questions whether a more clever mechanism exists to differentiate failed tasks from successful ones to prevent unfair resource consumption.