app.py

# AUTOGENERATED! DO NOT EDIT! File to edit: app.ipynb.

# %% auto 0 
__all__ = ['block', 'make_clickable_model', 'make_clickable_user', 'get_submissions']

# %% app.ipynb 0
import subprocess
import sys
def upgrade(package):
    subprocess.run([sys.executable, "-m", "pip", "install", "--upgrade", package])

upgrade("gradio")


import gradio as gr
import pandas as pd
from huggingface_hub import list_models

from diffusers import StableDiffusionPipeline

# %% app.ipynb 1
def get_model_list(category):
    submissions_list = list_models(filter=["dreambooth-hackathon", category], full=True)
    spaces_pipeline_load = [submission.id for submission in submissions_list ]
    return gr.Dropdown.update(choices=spaces_pipeline_load , value=spaces_pipeline_load[4])


def get_initial_prompt(model_nm):
    #Example - a photo of shbrcky dog 
    user_model_nm = model_nm.split('/')[-1]
    if '-' in user_model_nm:
        prompt = " ".join(user_model_nm.split('-'))
    else:
        prompt = user_model_nm
    return gr.Textbox.update(value="a photo of " + prompt + " ")

def get_pipeline(model_name): #, progress=gr.Progress(track_tqdm=True)):
    #Using diffusers pipeline to generate an image for the demo
    #Loading Your Dreambooth model
    pipeline = StableDiffusionPipeline.from_pretrained(model_name) # Example - ("ashiqabdulkhader/shiba-dog") or ('pharma/sugar-glider')
    return pipeline 

def make_demo(model_name, prompt, progress=gr.Progress(track_tqdm=True)):
    #Using diffusers pipeline to generate an image for the demo
    progress(0, desc="Starting...")
    pipeline = get_pipeline(model_name) #StableDiffusionPipeline.from_pretrained(model_name) # Example - ("ashiqabdulkhader/shiba-dog") or ('pharma/sugar-glider')
    #Generating Image from your prompt
    image_demo = pipeline(prompt).images[0]    
    return image_demo 

def make_clickable_model(model_name, link=None):
    if link is None:
        link = "https://huggingface.co/" + model_name
    # Remove user from model name
    return f'<a target="_blank" href="{link}">{model_name.split("/")[-1]}</a>'
    
def make_clickable_user(user_id):
    link = "https://huggingface.co/" + user_id
    return f'<a  target="_blank" href="{link}">{user_id}</a>'

# %% app.ipynb 2
def get_submissions(category, prompt):
    submissions = list_models(filter=["dreambooth-hackathon", category], full=True)
    leaderboard_models = []

    for submission in submissions:
        # user, model, likes
        user_id = submission.id.split("/")[0]
        model_nm = submission.id.split("/")[-1]
        if '-' in model_nm:
            model_nm = " ".join(model_nm.split('-'))
        #button_html = get_button()    
        leaderboard_models.append(
            (
                make_clickable_user(user_id),
                make_clickable_model(submission.id, prompt),
                submission.likes,
                #button_html #'a photo of ' + model_nm + " "
            )
        )

    df = pd.DataFrame(data=leaderboard_models, columns=["User", "Model", "Likes", ])
    df.sort_values(by=["Likes"], ascending=False, inplace=True)
    df.insert(0, "Rank", list(range(1, len(df) + 1)))
    return df

# %% app.ipynb 3
block = gr.Blocks()

with block:
    gr.Markdown(
        """# Gradio-powered leaderboard-evaluator for the DreamBooth Hackathon
    
    Welcome to this Gradio-powered leaderboard! Select a theme and one of the dreambooth models trained by hackathon-participants, and key in your prompt as shown (eg., a photo of Shiba dog in a jungle). Note that, the image generation might take long (around 400 seconds) as it will have to load the respective model pipeline into memory.  
    <br>**If you like a model demo, click on the model name in the table below and UPVOTE the model on Huggingface hub**<br><br>
    DreamBooth Hackathon - is an ongoing community event where participants **personalize a Stable Diffusion model** by fine-tuning it with a powerful technique called [_DreamBooth_](https://arxiv.org/abs/2208.12242). This technique allows one to implant a subject into the output domain of the model such that it can be synthesized with a _unique identifier_ (eg., shiba dog) in the prompt. 
        This competition comprises 5 _themes_ - Animals, Science, Food, Landscapes, and Wildcards. For details on how to participate, check out the hackathon's guide [here](https://github.com/huggingface/diffusion-models-class/blob/main/hackathon/README.md).
    """
    )
    with gr.Row():
        with gr.Column():
            theme = gr.Radio(label="Pick a Theme",choices=["animal","science", "food", "landscape", "wildcard"] )
            model_list = gr.Dropdown(label="Pick a Dreamboooth model", choices = []) # choices=
        with gr.Column():
            prompt_in = gr.Textbox(label="Type in a Prompt in front of the given text..", value="")
            button_in = gr.Button(Value = "Generate Image")
        image_out = gr.Image(label="Generated image with your choice of Dreambooth model")
    with gr.Tabs():
        with gr.TabItem("Animal 🐨"):
            with gr.Row():
                animal_data = gr.components.Dataframe(
                    type="pandas", datatype=["number", "markdown", "markdown", "number","str"], interactive = True
                )
            with gr.Row():
                data_run = gr.Button("Refresh")
                data_run.click(
                    get_submissions, inputs=[gr.Variable("animal"), prompt_in], outputs=animal_data
                )

                
        with gr.TabItem("Science 🔬"):
            with gr.Row():
                science_data = gr.components.Dataframe(
                    type="pandas", datatype=["number", "markdown", "markdown", "number", "str"], interactive = True
                )
            with gr.Row():
                data_run = gr.Button("Refresh")
                data_run.click(
                    get_submissions, inputs=[gr.Variable("science"), prompt_in], outputs=science_data
                )
        with gr.TabItem("Food 🍔"):
            with gr.Row():
                food_data = gr.components.Dataframe(
                    type="pandas", datatype=["number", "markdown", "markdown", "number", "str"], interactive = True
                )
            with gr.Row():
                data_run = gr.Button("Refresh")
                data_run.click(
                    get_submissions, inputs=[gr.Variable("food"), prompt_in], outputs=food_data
                )
        with gr.TabItem("Landscape 🏔"):
            with gr.Row():
                landscape_data = gr.components.Dataframe(
                    type="pandas", datatype=["number", "markdown", "markdown", "number", "str"], interactive = True
                )
            with gr.Row():
                data_run = gr.Button("Refresh")
                data_run.click(
                    get_submissions,
                    inputs=[gr.Variable("landscape"),prompt_in],
                    outputs=landscape_data,
                )
        with gr.TabItem("Wilcard 🔥"):
            with gr.Row():
                wildcard_data = gr.components.Dataframe(
                    type="pandas", datatype=["number", "markdown", "markdown", "number", "str"], interactive = True
                )
            with gr.Row():
                data_run = gr.Button("Refresh")
                data_run.click(
                    get_submissions,
                    inputs=[gr.Variable("wildcard"),prompt_in],
                    outputs=wildcard_data,
                )
    
    theme.change(get_model_list, theme, model_list )  
    model_list.change(get_initial_prompt, model_list, prompt_in )
    button_in.click(make_demo, [model_list, prompt_in], image_out)
    
    block.load(get_submissions, inputs=[gr.Variable("animal"), prompt_in], outputs=animal_data)
    block.load(get_submissions, inputs=[gr.Variable("science"), prompt_in], outputs=science_data)
    block.load(get_submissions, inputs=[gr.Variable("food"), prompt_in], outputs=food_data)
    block.load(get_submissions, inputs=[gr.Variable("landscape"), prompt_in], outputs=landscape_data)
    block.load(get_submissions, inputs=[gr.Variable("wildcard"), prompt_in], outputs=wildcard_data)

block.queue(concurrency_count=3)
block.launch()