import gradio as gr
from transformers import BlipForConditionalGeneration, AutoProcessor
from PIL import Image
import torch

# Load model and processor
processor = AutoProcessor.from_pretrained("blip-fine-tuned/")
processor.tokenizer.padding_size = 'left'
model = BlipForConditionalGeneration.from_pretrained("blip-fine-tuned/")
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)

def predict(image):
    # Preprocess the image
    inputs = processor(images=image, return_tensors="pt").to(device)
    pixel_values = inputs.pixel_values

    # get predictions
    with torch.no_grad():
        generated_ids = model.generate(pixel_values=pixel_values, max_length=100)
    
    generated_caption = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
    return generated_caption


# interface = gr.Interface(fn=predict, inputs=gr.Image(type="pil"), outputs="text")
interface = gr.Interface(
    fn=predict, 
    inputs=gr.Image(type="pil"), 
    outputs="text",
    title="BLIP Image Caption Generator",
    description="Upload an image or select a sample to generate a descriptive caption."  # Add description here
)
interface.launch()