[FEATURE] write tutorial on optimizing `HiDream-ai/HiDream-I1-Dev`

[davidberenstein1957]

Is your feature request related to a problem? Please describe.

Create an end-to-end tutorial that is tangible and showcases real-world examples! A good example is: https://docs.pruna.ai/en/stable/setup/replicate.html.
Some suggestions, but feel free to propose anything you think is cool.

• Title:
• Model: https://huggingface.co/HiDream-ai/HiDream-I1-Fast (perhaps use dev if compute allows for it?)
• Dataset prompts to generate examples from: https://huggingface.co/datasets/data-is-better-together/open-image-preferences-v1-binarized

Steps:

• Introduce outline
• Install Pruna
• Introduce and optimize the model and push the model to Hugging Face
• Explore data (embedded viewer)
• Evaluate the model using data
• Deploy as Gradio ZeroGPU Space / FastAPI

Describe the solution you'd like

Additional context

[ParagEkbote]

Thanks for raising this issue.

I had a couple of questions:

What should be the ideal size for the training dataset and which framework can be used for evaluation ?

cc: @davidberenstein1957

[davidberenstein1957]

Hi @ParagEkbote, we don't actually train the models, although they could be used for calibration for some algorithms. I think we can use this dataset for that, as well as the evaluation.

I recently wrote new docs (still need to be deployed) but I think it might help you understand the framework better and provide a nice evaluation overview: https://docs.pruna.ai/en/docs-general-structure-refactor/docs_pruna/user_manual/evaluate.html

[ParagEkbote]

Hi @ParagEkbote, we don't actually train the models, although they could be used for calibration for some algorithms. I think we can use this dataset for that, as well as the evaluation.

I recently wrote new docs (still need to be deployed) but I think it might help you understand the framework better and provide a nice evaluation overview: https://docs.pruna.ai/en/docs-general-structure-refactor/docs_pruna/user_manual/evaluate.html

You have shared a private link which I'm unable to view. Could you please share some other relevant resources?

[davidberenstein1957]

@ParagEkbote I made the documentation public.

Hi @ParagEkbote, we don't actually train the models, although they could be used for calibration for some algorithms. I think we can use this dataset for that, as well as the evaluation.
I recently wrote new docs (still need to be deployed) but I think it might help you understand the framework better and provide a nice evaluation overview: https://docs.pruna.ai/en/docs-general-structure-refactor/docs_pruna/user_manual/evaluate.html

You have shared a private link which I'm unable to view. Could you please share some other relevant resources?