Modularize InstructPix2Pix SDXL inferencing during and after training in examples

[sangyeon-k]

What does this PR do?

Partially fixes #6545 regarding InstructPix2Pix SDXL.

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline?
• Did you read our philosophy doc (important for complex PRs)?
• Was this discussed/approved via a GitHub issue or the forum? Please add a link to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?

Who can review?

@sayakpaul

[sayakpaul]

I think the PR is not yet ready (conflict).

[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.

[sangyeon-k]

@sayakpaul Thanks for letting me know.
I resolved the conflict and I think it is ready for review now.

[sayakpaul]

Looks cool. Do you have a training command for me to test this with?

Do the changes work as expected?

[sangyeon-k]

@sayakpaul Yes, it works as expected.
Here is the sequence of validation images with the prompt, "make it in japan" :)
When making the GIF file, I resized the images to 256x256 to comply with the upload size limitations.

Original	Edited

Regarding the training command, I used the one below.

export DATASET_ID="fusing/instructpix2pix-1000-samples"

accelerate launch train_instruct_pix2pix_sdxl.py \
    --pretrained_model_name_or_path=stabilityai/stable-diffusion-xl-base-1.0 \
    --pretrained_vae_model_name_or_path=madebyollin/sdxl-vae-fp16-fix \
    --dataset_name=$DATASET_ID \
    --use_ema \
    --enable_xformers_memory_efficient_attention \
    --resolution=512 --random_flip \
    --train_batch_size=4 --gradient_accumulation_steps=4 --gradient_checkpointing \
    --max_train_steps=15000 \
    --checkpointing_steps=5000 --checkpoints_total_limit=1 \
    --learning_rate=5e-05 --lr_warmup_steps=0 \
    --conditioning_dropout_prob=0.05 \
    --seed=42 \
    --val_image_url_or_path="https://datasets-server.huggingface.co/assets/fusing/instructpix2pix-1000-samples/--/fusing--instructpix2pix-1000-samples/train/23/input_image/image.jpg" \
    --validation_prompt="make it in japan" \
    --report_to=wandb \
    --push_to_hub \

[sayakpaul]

Lovely. I am gonna go ahead and merge. Thanks so much for this valuable contribution.