#from transformers import BlenderbotTokenizer, BlenderbotForConditionalGeneration
from transformers import AutoModelForCausalLM, AutoTokenizer,BlenderbotForConditionalGeneration
import torch


import torch

chat_tkn = AutoTokenizer.from_pretrained("facebook/blenderbot-400M-distill")
mdl = BlenderbotForConditionalGeneration.from_pretrained("facebook/blenderbot-400M-distill")

def converse(user_input, chat_history=[]):
    
    user_input_ids = chat_tkn.encode(user_input + chat_tkn.eos_token, return_tensors='pt')

    # create a combined tensor with chat history
    bot_input_ids = torch.cat([torch.LongTensor(chat_history), user_input_ids], dim=-1)

    # generate a response 
    chat_history = mdl.generate(bot_input_ids, max_length=1000, pad_token_id=chat_tkn.eos_token_id).tolist()
    print (chat_history)

    # convert the tokens to text, and then split the responses into lines
    response = chat_tkn.batch_decode(chat_history[0],skip_special_tokens=True)
    #response.remove("")
    print("starting to print response")
    print(response)
    
    # write some HTML
    html = "<div class='chatbot'>"
    for m, msg in enumerate(response):
        cls = "user" if m%2 == 0 else "bot"
        print("value of m")
        print(m)
        print("message")
        print (msg)
        html += "<div class='msg {}'> {}</div>".format(cls, msg)
    html += "</div>"
    print(html)
    return html, chat_history

import gradio as gr

css = """
.chatbox {display:flex;flex-direction:column}
.msg {padding:4px;margin-bottom:4px;border-radius:4px;width:80%}
.msg.user {background-color:cornflowerblue;color:white}
.msg.bot {background-color:lightgray;align-self:self-end}
.footer {display:none !important}
"""

gr.Interface(fn=converse,
             theme="default",
             inputs=[gr.inputs.Textbox(placeholder="How are you?"), "state"],
             outputs=["html", "state"],
             css=css).launch()