app.py

import torch
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
from PIL import Image
import gradio as gr

# Load text-to-image model
text_to_image_model = torch.hub.load("ProPerNounpYK/texttoimage", map_location=torch.device("cuda" if torch.cuda.is_available() else "cpu"))

# Load chat model
chat_model = AutoModelForSeq2SeqLM.from_pretrained("ProPerNounpYK/chat")
chat_tokenizer = AutoTokenizer.from_pretrained("ProPerNounpYK/chat")

# Create multimodal interface
interface = gr.Interface(
    fn=lambda input_text, input_image: generate_response(input_text, input_image),
    inputs=["text", "image"],
    outputs=["text", "image"],
    title="Multimodal Conversational AI",
    description="Talk to me, and I'll respond with images!"
)

def generate_response(input_text, input_image):
    # Process input text using chat model
    chat_output = chat_model(input_text)
    chat_response = chat_tokenizer.decode(chat_output, skip_special_tokens=True)

    # Process input image using text-to-image model
    generated_image = text_to_image_model(input_text, input_image)

    # Return response as a tuple of text and image
    return chat_response, generated_image

# Launch the interface
interface.launch()