import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# 加载模型和分词器
model_name = "jinaai/reader-lm-1.5b"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.bfloat16).to("cuda")

# 定义转换函数
def html_to_markdown(html_content):
    messages = [{"role": "user", "content": html_content}]
    input_text = tokenizer.apply_chat_template(messages, tokenize=False)
    inputs = tokenizer.encode(input_text, return_tensors="pt").to("cuda")
    outputs = model.generate(inputs, max_new_tokens=1024, temperature=0, do_sample=False, repetition_penalty=1.08)
    markdown_content = tokenizer.decode(outputs[0], skip_special_tokens=True)
    return markdown_content

# 创建 Gradio 接口
iface = gr.Interface(
    fn=html_to_markdown,
    inputs=gr.inputs.Textbox(lines=20, label="HTML 内容"),
    outputs=gr.outputs.Textbox(label="Markdown 内容"),
    title="HTML 转 Markdown 转换器",
    description="输入 HTML 内容，模型将其转换为 Markdown 格式。",
)

if __name__ == "__main__":
    iface.launch()