When using the qwen-image model to generate images, diffusers takes twice as long as diffsynth-studio

[gelinhua]

diffsynth-studio took 9 minutes and 58 seconds diffusrs took 17 minutes and 40 seconds

The prompt words, number of steps, length and width are all consistent

[SahilCarterr]

Can you share the inference Code @gelinhua

[asomoza]

simple guess is that one its using CFG and the other not. The pipeline it's very peculiar about this, if you don't provide a negative prompt it doesn't use CFG even if the true_cfg is more than 1.0

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.