app.py

import gradio as gr
from PIL import Image
import requests
import torch
from transformers import BlipForConditionalGeneration, BlipProcessor, GenerationConfig

device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')


_MODEL_PATH = 'IDEA-CCNL/Taiyi-BLIP-750M-Chinese'

processor = BlipProcessor.from_pretrained(_MODEL_PATH)
model = BlipForConditionalGeneration.from_pretrained(_MODEL_PATH).half().eval().to(device)


def inference(raw_image, model_n, question, strategy):
    if model_n == 'Image Captioning':
        image = processor(raw_image).to(device, torch.float16)
        with torch.no_grad():
            if strategy == "Beam search":
                config = GenerationConfig(
                    do_sample=False,
                    num_beams=3,
                    max_length=20,
                    min_length=5,
                )
                captions = model.generate(image, generation_config=config)
            else:
                config = GenerationConfig(
                    do_sample=True,
                    top_p=0.9,
                    max_length=20,
                    min_length=5,
                )
                captions = model.generate(image, generation_config=config)
            caption = processor.decode(captions[0], skip_special_tokens=True)
            caption = caption.replace(' ', '')
            return 'caption: '+caption


inputs = [gr.inputs.Image(type='pil'), gr.inputs.Radio(choices=['Image Captioning'], type="value", default="Image Captioning", label="Task"), gr.inputs.Textbox(
    lines=2, label="Question"), gr.inputs.Radio(choices=['Beam search', 'Nucleus sampling'], type="value", default="Nucleus sampling", label="Caption Decoding Strategy")]
outputs = gr.outputs.Textbox(label="Output")

title = "BLIP"

description = "Gradio demo for BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation (Salesforce Research). To use it, simply upload your image, or click one of the examples to load them. Read more at the links below."

article = "<p style='text-align: center'><a href='https://github.com/IDEA-CCNL/Fengshenbang-LM' target='_blank'>Github Repo</a></p>"


gr.Interface(inference, inputs, outputs, title=title, description=description, article=article, examples=[
             ['starrynight.jpeg', "Image Captioning", "None", "Nucleus sampling"]]).launch(enable_queue=True)