app.py

import gradio as gr
from transformers import AutoProcessor, AutoModelForCausalLM
import spaces
import re
from PIL import Image 

import subprocess
subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)

model = AutoModelForCausalLM.from_pretrained('gokaygokay/Florence-2-SD3-Captioner', trust_remote_code=True).to("cuda").eval()

processor = AutoProcessor.from_pretrained('gokaygokay/Florence-2-SD3-Captioner', trust_remote_code=True)


TITLE = "# [Florence-2 SD3 Long Captioner](https://huggingface.co/gokaygokay/Florence-2-SD3-Captioner/)"
DESCRIPTION = "[Florence-2 Base](https://huggingface.co/microsoft/Florence-2-base-ft) fine-tuned on Long SD3 Prompt and Image pairs. Check above link for datasets that are used for fine-tuning."

def modify_caption(caption: str) -> str:
    """
    Removes specific prefixes from captions if present, otherwise returns the original caption.
    Args:
        caption (str): A string containing a caption.
    Returns:
        str: The caption with the prefix removed if it was present, or the original caption.
    """
    # Define the prefixes to remove
    prefix_substrings = [
        ('captured from ', ''),
        ('captured at ', '')
    ]
    
    # Create a regex pattern to match any of the prefixes
    pattern = '|'.join([re.escape(opening) for opening, _ in prefix_substrings])
    replacers = {opening.lower(): replacer for opening, replacer in prefix_substrings}
    
    # Function to replace matched prefix with its corresponding replacement
    def replace_fn(match):
        return replacers[match.group(0).lower()]
    
    # Apply the regex to the caption
    modified_caption = re.sub(pattern, replace_fn, caption, count=1, flags=re.IGNORECASE)
    
    # If the caption was modified, return the modified version; otherwise, return the original
    return modified_caption if modified_caption != caption else caption
    
@spaces.GPU
def run_example(image):
    image = Image.fromarray(image)
    task_prompt = "<DESCRIPTION>"
    prompt = task_prompt + "Describe this image in great detail."

    # Ensure the image is in RGB mode
    if image.mode != "RGB":
        image = image.convert("RGB")

    inputs = processor(text=prompt, images=image, return_tensors="pt").to("cuda")
    generated_ids = model.generate(
        input_ids=inputs["input_ids"],
        pixel_values=inputs["pixel_values"],
        max_new_tokens=1024,
        num_beams=3
    )
    generated_text = processor.batch_decode(generated_ids, skip_special_tokens=False)[0]
    parsed_answer = processor.post_process_generation(generated_text, task=task_prompt, image_size=(image.width, image.height))
    return modify_caption(parsed_answer["<DESCRIPTION>"])


css = """
  #output {
    height: 500px; 
    overflow: auto; 
    border: 1px solid #ccc; 
  }
"""

with gr.Blocks(css=css) as demo:
    gr.Markdown(TITLE)
    gr.Markdown(DESCRIPTION)
    with gr.Tab(label="Florence-2 SD3 Prompts"):
        with gr.Row():
            with gr.Column():
                input_img = gr.Image(label="Input Picture")
                submit_btn = gr.Button(value="Submit")
            with gr.Column():
                output_text = gr.Textbox(label="Output Text")

        gr.Examples(
            [["image1.jpg"], ["image2.jpg"], ["image3.png"], ["image4.jpg"], ["image5.jpg"], ["image6.PNG"]],
            inputs = [input_img],
            outputs = [output_text],
            fn=run_example,
            label='Try captioning on below examples'
            )

        submit_btn.click(run_example, [input_img], [output_text])

demo.launch(debug=True)