import gradio as gr
import spaces
import torch
import os
import subprocess

os.system("pip install git+https://github.com/huggingface/transformers")

from transformers import Qwen2VLForConditionalGeneration, AutoTokenizer, AutoProcessor
from qwen_vl_utils import process_vision_info

model = Qwen2VLForConditionalGeneration.from_pretrained("Qwen/Qwen2-VL-7B-Instruct").cuda()
processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-7B-Instruct")

@spaces.GPU
def infer(u, t):
    if len(u) < 1:
        u = "https://lf3-static.bytednsdoc.com/obj/eden-cn/pbovhozuha/screenshot-20240923-164458.png"
    if len(t) < 1:
        t = "请将图里文字转成markdown"
    messages = [
        {
            "role": "user",
            "content": [
                {
                    "type": "image",
                    "image": u,
                },
                {
                    "type": "text",
                    "text": t,
                },
            ],
        }
    ]
    text = processor.apply_chat_template(
        messages, tokenize=False, add_generation_prompt=True
    )
    image_inputs, video_inputs = process_vision_info(messages)
    inputs = processor(
        text=[text],
        images=image_inputs,
        videos=video_inputs,
        padding=True,
        return_tensors="pt",
    )
    inputs = inputs.to(model.device)
    generated_ids = model.generate(**inputs, max_new_tokens=512)
    generated_ids_trimmed = [
        out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
    ]
    output_text = processor.batch_decode(
        generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
    )
    return output_text
    

demo = gr.Interface(
    fn=infer,
    inputs=[
        gr.Text(label="url"),
        gr.Text(label="text"),
    ],
    outputs=gr.Text(),
)
demo.launch()