from transformers import VisionEncoderDecoderModel, ViTImageProcessor, BertTokenizer
import torch
from PIL import Image

# load models
feature_extractor = ViTImageProcessor.from_pretrained("atasoglu/vit-bert-flickr8k")
tokenizer = BertTokenizer.from_pretrained("atasoglu/vit-bert-flickr8k")
model = VisionEncoderDecoderModel.from_pretrained("atasoglu/vit-bert-flickr8k")

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)

# load image
img = Image.open("example.jpg")

# encode (extracting features)
pixel_values = feature_extractor(images=[img], return_tensors="pt").pixel_values
pixel_values = pixel_values.to(device)

# generate caption
output_ids = model.generate(pixel_values)

# decode
preds = tokenizer.batch_decode(output_ids, skip_special_tokens=True)
print(preds)