What's the difference between the batch count and batch size options?

[UntouchedWagons]

When would I opt for a larger batch count over a larger batch size or vice versa?

[Woolverine94]

Hello @UntouchedWagons,

Thanks for you question and you interest in the project.

Difference is the same that between parallel and serial processing :

• Batch size = number of contents to generate in a single inference, using the same settings (parallel)
• Batch count = number of inferences to repeat automatically, using the same settings (serial)

Batch size is faster than batch count for an equal value, but take more resources (RAM/VRAM mainly).

Both are cumulative and you can, for example, run 4 batch count of 4 batch size to obtain 16 generated contents.

Unless you have limited resources, there's not a real reason to prefer batch count over batch size.

Also note that on most modules, batch size will use incremental seeds in the same inference, which is not the case with batch count that will use random seed for all inferences.

Hoping this will answer your question.

[Woolverine94]

I correct myself : batch count also use incremental seeds.