import torch
from PIL import Image
from transformers import BlipProcessor, BlipForConditionalGeneration
import gradio as gr
from gradio.themes.utils import sizes


theme = gr.themes.Default(radius_size=sizes.radius_none).set(
    block_label_text_color = '#4D63FF',
    block_title_text_color = '#4D63FF',
    button_primary_text_color = '#4D63FF',
    button_primary_background_fill='#FFFFFF',
    button_primary_border_color='#4D63FF',
    button_primary_background_fill_hover='#EDEFFF',
)

processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large", torch_dtype=torch.float16).to("cuda")


def image2text(image):
    inp = Image.fromarray(image.astype('uint8'), 'RGB')
    text = "a photography of"
    inputs = processor(inp, text, return_tensors="pt").to("cuda", torch.float16)
    out = model.generate(**inputs)

    return processor.decode(out[0], skip_special_tokens=True)

with gr.Blocks(theme=theme, css="footer {visibility: hidden}") as demo:
    gr.Markdown("""
    <div align='center' ><font size='60'>照片描述</font></div>
    """)
    with gr.Row():
        with gr.Column():
            box1 = gr.Image(label="图片")
            with gr.Row():
                button = gr.Button("提交", variant="primary")
        box2 = gr.Textbox(label="文本")

    button.click(fn=image2text, inputs=box1, outputs=box2)
    examples = gr.Examples(examples=[['soccer.jpg']],  inputs=[box1], label="例子")


if __name__ == "__main__":
    demo.queue(concurrency_count=3).launch(server_name = "0.0.0.0", server_port = 7018)