app.py

import gradio as gr
from transformers import pipeline

# Load model (dont really know what this is doing under the hood)
generator = pipeline('text-generation', model='pere/norwegian-gpt2-social')


# Create function that outputs generated text
def complete_prompt(prompt):
    output = generator(prompt, max_length=30, num_return_sequences=5)
    return "\n \n".join([l['generated_text'] for l in output])

demo = gr.Interface(fn=complete_prompt, inputs="text", outputs="text", examples=["Hvor var du når oddvar","god dag mann, "])

with gr.Blocks() as demo:
    gr.Markdown(
    """
    # Hello AI Academy!
    Ive made a very simple sentence completion app based on the norwegian-gpt2-social model on huggingface made by the user pere.
    It is trained on wikipedia, twitter, some news corpus, reddit and even VG debatt(!!), so prepare to be morally shocked by its results (I take no responsibility).

    """)
    gr.Interface(fn=complete_prompt, inputs="text", outputs="text", examples=["Hvor var du når oddvar","god dag mann, ", "moderatoren er", "svensker", "The worst thing about norway"])

if __name__ == "__main__":
    demo.launch()