[Sana Sprint] add image-to-image pipeline

[linoytsaban]

This PR adds sana sprint image-to-image ⚡️⚡️
@sayakpaul @apolinario

import torch
from diffusers import SanaSprintImg2ImgPipeline
from diffusers.utils.loading_utils import load_image

image = load_image(
    "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/penguin.png"
)

pipe = SanaSprintImg2ImgPipeline.from_pretrained("Efficient-Large-Model/Sana_Sprint_1.6B_1024px_diffusers", torch_dtype=torch.bfloat16)
pipe.to("cuda")

pipe(prompt="a cute pink bear", 
        image=image, 
        strength=0.45, 
        height=832, width=480).images[0]

run time on H100 -

100%
 1/1 [00:00<00:00, 14.24it/s]

[linoytsaban]

@bot /style

[github-actions]

Style fixes have been applied. View the workflow run here.

[sayakpaul]

@linoytsaban could you also update the PR with example code and results? And maybe also comment a bit on the speed repercussions?

[sayakpaul]

Thanks! Left some comments. Also, I think we need to add to docs.

[linoytsaban]

@bot /style

[github-actions]

Style fixes have been applied. View the workflow run here.

[sayakpaul]

Cc: @asomoza for testing (if you want to).

[sayakpaul]

Good to merge after adding a test.

[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.