import gradio as gr
from transformers import pipeline

# Loading the 124M version for lightning speed on CPU
generator = pipeline('text-generation', model='gpt2')

def predict(prompt):
    # GPT-2 needs a bit of 'length' to show its true colors
    output = generator(prompt, max_length=150, num_return_sequences=1)
    return output[0]['generated_text']

gr.Interface(fn=predict, inputs="text", outputs="text").launch()