import transformers
import gradio as gr
import git
import os
os.system("pip install --upgrade pip")

#Load arabert preprocessor
import git
git.Git("arabert").clone("https://github.com/aub-mind/arabert")
from arabert.preprocess import ArabertPreprocessor
arabert_prep = ArabertPreprocessor(model_name="bert-base-arabert", keep_emojis=False)


#Load Model
from transformers import EncoderDecoderModel, AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained("tareknaous/bert2bert-empathetic-response-msa")
model = EncoderDecoderModel.from_pretrained("tareknaous/bert2bert-empathetic-response-msa")
model.eval()

def generate_response(text, minimum_length, p, temperature):
  text_clean = arabert_prep.preprocess(text)
  inputs = tokenizer.encode_plus(text_clean,return_tensors='pt')
  outputs = model.generate(input_ids = inputs.input_ids,
                   attention_mask = inputs.attention_mask,
                   do_sample = True,
                   min_length=minimum_length,
                   top_p = p,
                   temperature = temperature)
  preds = tokenizer.batch_decode(outputs) 
  response = str(preds)
  response = response.replace("\'", '')
  response = response.replace("[[CLS]", '')
  response = response.replace("[SEP]]", '')
  response = str(arabert_prep.desegment(response))
  return response

# title = 'Empathetic Response Generation in Arabic'
# description = 'This demo is for a BERT2BERT model trained for single-turn open-domain empathetic dialogue response generation in Modern Standard Arabic'
css = """
.rtlClass  {direction:rtl !important}
"""
with gr.Blocks(css=css) as demo:
    with gr.Column():
        gr.Markdown("Empathetic Response Generation in Arabic")
        chatbot = gr.Chatbot(elem_classes="rtlClass").style(height=400)
        msg = gr.Textbox(placeholder="ارسل رسالة",show_label=False,elem_classes="rtlClass").style(container=False)
    with gr.Column():
        output_slider=gr.Slider(5, 20, step=1, label='Minimum Output Length')
        top_p_slider=gr.Slider(0.7, 1, step=0.1,  label='Top-P')
        temperature_slider=gr.Slider(1, 3, step=0.1, label='Temperature')
        def respond(message,chat_history,output_slider,top_p_slider,temperature_slider):
            bot_message = generate_response(message,output_slider,top_p_slider,temperature_slider)
            chat_history.append((message, bot_message))
            return "", chat_history
        clear = gr.Button("Clear Chat")
    msg.submit(respond, [msg, chatbot,output_slider,top_p_slider,temperature_slider], [msg, chatbot])
    clear.click(lambda: None, None, chatbot, queue=False)

demo.launch()