app.py

import gradio as gr
import re
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

# Modell und Tokenizer laden

# gpt2 outputs text!
#tokenizer = AutoTokenizer.from_pretrained("openai-community/gpt2")
#model = AutoModelForCausalLM.from_pretrained("openai-community/gpt2")

# distilgpt2 is only 80MB
tokenizer = AutoTokenizer.from_pretrained("distilgpt2")
model = AutoModelForCausalLM.from_pretrained("distilgpt2")


def text_to_emoji(text):
    # remove special characters
    text_cleaned = re.sub(r"[.,!?;:]", "", text)
    prompt = f"Convert the following sentence into an emoji-sequence which conveys a similar meaning and return only the emojis, no explanation:\n\n\"{text_cleaned}\""

    # Tokenisieren
    inputs = tokenizer(prompt, return_tensors="pt")
    
    # Antwort generieren
    outputs = model.generate(**inputs, max_new_tokens=25, do_sample=True, temperature=0.7)
    
    # Antwort decodieren
    result = tokenizer.decode(outputs[0], skip_special_tokens=True)
    
    return result

# Gradio UI
iface = gr.Interface(
    fn=text_to_emoji,
    inputs=gr.Textbox(lines=2, placeholder="Enter a sentence..."),
    outputs="text",
    title="AI-Powered Emoji Translator",
    description="Enter a sentence, and the AI will transform it into an emoji-version 🥳"
)

iface.launch()