app.py

import gradio as gr
from transformers import AutoTokenizer, AutoModelForCausalLM
import spaces
from PIL import Image
import io


models = {
    "jinaai/reader-lm-0.5b": AutoModelForCausalLM.from_pretrained("jinaai/reader-lm-0.5b", trust_remote_code=True).to("cuda").eval(),
}

tokenizers = {
    "jinaai/reader-lm-0.5b": AutoTokenizer.from_pretrained("jinaai/reader-lm-0.5b", trust_remote_code=True),
}


@spaces.GPU
def run_example(html_content, model_id="jinaai/reader-lm-0.5b"):
    model = models[model_id]
    tokenizer = tokenizers[model_id]
    messages = [{"role": "user", "content": html_content}]
    input_text=tokenizer.apply_chat_template(messages, tokenize=False)
    inputs = tokenizer.encode(input_text, return_tensors="pt").to(device)
    outputs = model.generate(inputs, max_new_tokens=1024, temperature=0, do_sample=False, repetition_penalty=1.08)
    return tokenizer.decode(outputs[0])


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

with gr.Blocks(css=css) as demo:
    gr.Markdown("""
    # HTML-to-Markdown
    """)
    with gr.Tab(label="Main"):
        with gr.Row():
            with gr.Column():
                model_selector = gr.Dropdown(choices=list(models.keys()), label="Model", value="maxiw/Florence-2-ScreenQA-base")
                html_content = gr.Textbox(label="HTML")
                submit_btn = gr.Button(value="Submit")
            with gr.Column():
                output_text = gr.Textbox(label="Markdown")

        submit_btn.click(run_example, [html_content, model_selector], [output_text])

demo.launch(debug=True)