import torch
import re
import gradio as gr
from pathlib import Path
from transformers import AutoTokenizer, AutoFeatureExtractor, VisionEncoderDecoderModel

def predict(image, max_length=30, num_beams=4):
    image = image.convert('RGB')
    pixel_values = feature_extractor(images=image, return_tensors="pt").pixel_values
    pixel_values = pixel_values.to(device)
    with torch.no_grad():
        caption_ids = model.generate(pixel_values.cpu())[0]
        caption_text = tokenizer.decode(caption_ids, skip_special_tokens=True)
    return caption_text
    
model_path = "MahsaShahidi/Persian-Image-Captioning"
device = "cpu"
# Load model.
model = VisionEncoderDecoderModel.from_pretrained(model_path)
model.to(device)
print("Loaded model")
feature_extractor = AutoFeatureExtractor.from_pretrained("google/vit-base-patch16-224-in21k")
print("Loaded feature_extractor")
tokenizer = AutoTokenizer.from_pretrained('HooshvareLab/bert-fa-base-uncased-clf-persiannews')
print("Loaded tokenizer")
title = "Persian Image Captioning"
description = ""

input = gr.inputs.Image(label="Image to search", type = 'pil', optional=False)
output = gr.outputs.Textbox(type="auto",label="Captions")

article = "This HuggingFace Space presents a demo for Persian Image Camptioning on VIT as its Encoder and ParsBERT (v2.0) as its Decoder"

images = [f"./image-{i}.jpg" for i in range(1,4)]

interface = gr.Interface(
        fn=predict,
        inputs = input,
        outputs=output,
        examples = images,
        title=title,
        description=article,
    )
interface.launch()