app.py

import gradio as gr
import torch
from transformers import Qwen2VLForConditionalGeneration, AutoProcessor
from qwen_vl_utils import process_vision_info
from PIL import Image
from datetime import datetime
import numpy as np
import os

# Function to save image array as a file and return the path
def array_to_image_path(image_array):
    img = Image.fromarray(np.uint8(image_array))
    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
    filename = f"image_{timestamp}.png"
    img.save(filename)
    return os.path.abspath(filename)

# Load model and processor
model = Qwen2VLForConditionalGeneration.from_pretrained(
    "Qwen/Qwen2-VL-2B-Instruct", 
    torch_dtype=torch.float32,
    device_map="cpu"
).eval()

processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-2B-Instruct")

DESCRIPTION = "[Qwen2-VL-2B Demo (CPU Version)](https://huggingface.co/Qwen/Qwen2-VL-2B-Instruct)"

def run_example(image, text_input):
    image_path = array_to_image_path(image)
    
    image = Image.fromarray(image).convert("RGB")
    messages = [
        {
            "role": "user",
            "content": [
                {
                    "type": "image",
                    "image": image_path,
                },
                {"type": "text", "text": text_input},
            ],
        }
    ]
    
    # Preparation for inference
    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",
    )
    
    # Inference: Generation of the output
    with torch.no_grad():
        generated_ids = model.generate(**inputs, max_new_tokens=128)
    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[0]

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

with gr.Blocks(css=css) as demo:
    gr.Markdown(DESCRIPTION)
    with gr.Tab(label="Qwen2-VL-2B Input (CPU)"):
        with gr.Row():
            with gr.Column():
                input_img = gr.Image(label="Input Picture")
                text_input = gr.Textbox(label="Question")
                submit_btn = gr.Button(value="Submit")
            with gr.Column():
                output_text = gr.Textbox(label="Output Text")

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

commandline_args = os.getenv("COMMANDLINE_ARGS", "")

# Enable or disable queue based on commandline_args
if "--no-gradio-queue" not in commandline_args:
    demo.queue(api_open=False)

demo.launch(inline=False, server_name="0.0.0.0", server_port=int(os.getenv("PORT", 7860)), debug=True)