caption_refiner.py

from transformers import T5Tokenizer, T5ForConditionalGeneration
import torch

# Load the T5 model and tokenizer
tokenizer = T5Tokenizer.from_pretrained("t5-base")
model = T5ForConditionalGeneration.from_pretrained("t5-base")

def refine_caption(initial_caption):
    # Prepare the input for the T5 model
    input_text = f"refine caption: {initial_caption}"
    inputs = tokenizer(input_text, return_tensors="pt", padding=True, truncation=True, max_length=512)
    
    # Generate the refined caption
    with torch.no_grad():
        outputs = model.generate(inputs.input_ids, max_new_tokens=100, num_return_sequences=1)
    
    # Decode the refined caption
    refined_caption = tokenizer.decode(outputs[0], skip_special_tokens=True)
    
    return refined_caption