app.py

import gradio as gr
from gradio import mix
import numpy as np
import torch
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
from huggingface_hub import from_pretrained_keras

tokenizer = AutoTokenizer.from_pretrained("bert-base-cased")
model = from_pretrained_keras("keras-io/text-generation-miniature-gpt")

def tokenize_data(text):
    # Tokenize the review body
    input_ =  str(text) + ' </s>'
    max_len = 80
    # tokenize inputs
    tokenized_inputs = tokenizer(input_, padding='max_length', truncation=True, max_length=max_len, return_attention_mask=True, return_tensors='pt')

    inputs={"input_ids": tokenized_inputs['input_ids'],
        "attention_mask": tokenized_inputs['attention_mask']}
    return inputs

def generate_answers(text):
    inputs = tokenize_data(text)
    x = tf.dtypes.cast(inputs['input_ids'], tf.int64)
    predictions, _ = model.predict(x)
    #predictions, _ = model.predict(inputs['input_ids'])
    results = np.argmax(predictions, axis=1)
    answer = tokenizer.decode(results[0].flatten(), skip_special_tokens=True)
    return answer

examples = [["Once upon a time, "], ["In a galaxy far far away"]]
title = "Text Generation with Miniature GPT"
description = "Gradio Demo for a miniature with GPT. To use it, simply add your text, or click one of the examples to load them. Read more at the links below."

iface = gr.Interface(fn=generate_answers, title = title, description=description, inputs=['text'], outputs=["text"], examples=examples)
iface.launch(inline=False, share=True)