Improve post-processing performance

[soof-golan]

What does this PR do?

Fixes: #10158

• Use multiplication instead of division in VaeImageProcessor.denormalize
• Avoid splitting and re-stacking tensors to reduce memory bandwidth and CPU-GPU syncs

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[hlky]

Thanks for your contribution @soof-golan! Can you run make style?

[soof-golan]

Thanks for your contribution @soof-golan! Can you run make style?

@hlky Done!

[yiyixuxu]

I think there is some context to that, each image in the batch may have a different value for do_normalize for sd1.5, see the code here

diffusers/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py

Line 1075 in 6131a93

image = self.image_processor.postprocess(image, output_type=output_type, do_denormalize=do_denormalize)

Since most of the new pipelines (sdxl, sd3, flux), we do not pass the do_normalize from the pipeline, i.e. do_normalize is None here , see SDXL https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl.py#L1304 ,

and we already did this so it will be batch processed for all the new pipelines already, I think this line is sufficient here

        if do_denormalize is None:
            return self.denormalize(images) if self.config.do_normalize else images

[yiyixuxu]

if do_denormalize is None:
            image = self.denormalize(images) if self.config.do_normalize 
else: 
        image = torch.stack(
            [self.denormalize(image[i]) if do_denormalize[i] else image[i] for i in range(image.shape[0])]
        )

[soof-golan]

@yiyixuxu I've opened up #10170 with only the simpler optimization, leaving the homogeneous case untouched.

Choose whichever you see fit for the scope of the project :)

[hlky]

Closed by #10170