app.py

import torch
device = "cuda:0" if torch.cuda.is_available() else "cpu"

from transformers import GPT2LMHeadModel, GPT2Tokenizer
import streamlit as st

st.set_page_config(
    page_title="GPT-2 Demo", 
    page_icon=":robot_face:", 
    layout="wide")

st.title("GPT-2 Text Generation Demo")
st.info("This is an GPT2 Text Generation Example using HuggingFace GPT2 Model")

pretrained = "gpt2-large"
tokenizer = GPT2Tokenizer.from_pretrained(pretrained)
model = GPT2LMHeadModel.from_pretrained(pretrained, pad_token_id=tokenizer.eos_token_id)

sentence = st.text_input('Input your sentence here:', value='My favorite ice cream flavor is ')

st.info("Max generated sentence: 100 words")
if (st.button("Generate")):
    input_ids = tokenizer.encode(sentence, return_tensors='pt').to(device)
    paragraph_generated = model.generate(input_ids, max_length=100, num_beams=5, no_repeat_ngram_size=2, early_stopping=True).to(device)
    text = tokenizer.decode(paragraph_generated[0], skip_special_tokens=True)

    st.write(text)