app.py

import numpy as np
import torch
import torchvision.transforms as T
from PIL import Image
from torchvision.transforms.functional import InterpolationMode
from transformers import AutoModel, AutoTokenizer
import gradio as gr

IMAGENET_MEAN = (0.485, 0.456, 0.406)
IMAGENET_STD = (0.229, 0.224, 0.225)

# Build the image transform
def build_transform(input_size):
    MEAN, STD = IMAGENET_MEAN, IMAGENET_STD
    transform = T.Compose([
        T.Lambda(lambda img: img.convert('RGB') if img.mode != 'RGB' else img),
        T.Resize((input_size, input_size), interpolation=InterpolationMode.BICUBIC),
        T.ToTensor(),
        T.Normalize(mean=MEAN, std=STD)
    ])
    return transform

# Dynamic preprocessing
def dynamic_preprocess(image, min_num=1, max_num=12, image_size=448, use_thumbnail=False):
    orig_width, orig_height = image.size
    aspect_ratio = orig_width / orig_height
    target_ratios = sorted(
        set((i, j) for n in range(min_num, max_num + 1) for i in range(1, n + 1) for j in range(1, n + 1) if i * j <= max_num and i * j >= min_num),
        key=lambda x: x[0] * x[1]
    )
    target_aspect_ratio = target_ratios[0]
    target_width = image_size * target_aspect_ratio[0]
    target_height = image_size * target_aspect_ratio[1]
    blocks = target_aspect_ratio[0] * target_aspect_ratio[1]
    resized_img = image.resize((target_width, target_height))
    processed_images = [
        resized_img.crop((
            (i % (target_width // image_size)) * image_size,
            (i // (target_width // image_size)) * image_size,
            ((i % (target_width // image_size)) + 1) * image_size,
            ((i // (target_width // image_size)) + 1) * image_size
        ))
        for i in range(blocks)
    ]
    if use_thumbnail and len(processed_images) != 1:
        thumbnail_img = image.resize((image_size, image_size))
        processed_images.append(thumbnail_img)
    return processed_images

# Load image dynamically from user upload
def load_image(image, input_size=448, max_num=12):
    transform = build_transform(input_size=input_size)
    images = dynamic_preprocess(image, image_size=input_size, use_thumbnail=True, max_num=max_num)
    pixel_values = [transform(image) for image in images]
    pixel_values = torch.stack(pixel_values)
    return pixel_values

# Load the model and tokenizer
path = 'OpenGVLab/InternVL2_5-1B'
model = AutoModel.from_pretrained(
    path,
    torch_dtype=torch.bfloat16,
    low_cpu_mem_usage=True,
    use_flash_attn=True,
    trust_remote_code=True
).eval().cuda()

tokenizer = AutoTokenizer.from_pretrained(path, trust_remote_code=True, use_fast=False)

# Define the function for Gradio interface
def process_image(image):
    try:
        pixel_values = load_image(image, max_num=12).to(torch.bfloat16).cuda()
        generation_config = dict(max_new_tokens=1024, do_sample=True)
        question = '<image>\nExtract text from the image, respond with only the extracted text.'
        response = model.chat(tokenizer, pixel_values, question, generation_config)
        return response
    except Exception as e:
        return f"Error: {str(e)}"

# Create Gradio Interface
demo = gr.Interface(
    fn=process_image,
    inputs=gr.Image(type="pil"),
    outputs="text",
    title="Dynamic Image Processing with InternVL",
    description="Upload an image and get detailed responses using the InternVL model."
)

# Launch the Gradio app
if __name__ == "__main__":
    demo.launch(server_name="0.0.0.0", server_port=7860)