import gradio as gr
from gradio.components import Slider, Textbox, Radio

import tensorflow as tf
from transformers import GPT2LMHeadModel, GPT2Tokenizer

global tokenizer, model, script_speaker_name, script_responder_name, convo
tokenizer = GPT2Tokenizer.from_pretrained("ethzanalytics/ai-msgbot-gpt2-XL-dialogue")
model = GPT2LMHeadModel.from_pretrained("ethzanalytics/ai-msgbot-gpt2-XL-dialogue", pad_token_id=tokenizer.eos_token_id)

script_speaker_name = "person alpha"

script_responder_name = "person beta"

global convo
convo = ""
def output(prompt, output_length):
    global convo
    if prompt.split(" ")[0].strip()=="clear_convo()":
        convo = ""
        prompt = prompt.split("clear_convo()")[1].strip()
    sentence = convo + '\n' + script_speaker_name + ': ' + prompt + '\n' + script_responder_name + ': '
    input_ids = tokenizer.encode(sentence, return_tensors='pt')
    # generate text until the output length (which includes the context length) reaches 50
    output = model.generate(input_ids, max_new_tokens=output_length, num_beams=5, no_repeat_ngram_size=2, early_stopping=True)
    convo = tokenizer.decode(output[0], skip_special_tokens=True)
    return convo
convo = ''

iface = gr.Interface(fn=output, inputs=["text", Slider(minimum=1.0, maximum=1000.0, step=1.0, default=50.0, label="Output Length")], outputs="text")
iface.launch()