app.py

import gradio as gr
import torch
from transformers import Qwen2VLForConditionalGeneration, AutoProcessor
from qwen_vl_utils import process_vision_info

# Load the model and processor on available device(s)
model = Qwen2VLForConditionalGeneration.from_pretrained(
    "Qwen/Qwen2-VL-72B-Instruct-AWQ",
    torch_dtype=torch.float16,
    #device_map="auto"
)

processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-72B-Instruct-AWQ")

@spaces.GPU(duration=60)
def generate_caption(image, prompt):
    messages = [
        {
            "role": "user",
            "content": [
                {
                    "type": "image",
                    "image": image,  # The uploaded image
                },
                {"type": "text", "text": prompt},
            ],
        }
    ]
   
    # Prepare the input
    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"
    )
    device = "cuda" if torch.cuda.is_available() else "cpu"
    inputs = inputs.to(device)

    # Generate the output
    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]


# Launch the Gradio interface with the updated inference function and title
demo = gr.ChatInterface(fn=generate_caption, title="Qwen2-VL-72B-Instruct-OCR", multimodal=True, description="Upload your Image and get the best possible insights out of the Image")
demo.queue().launch()