File size: 1,861 Bytes
e966862 3be58b8 e966862 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 |
# https://ai-brewery.medium.com/conversational-chatbot-using-transformers-and-streamlit-73d621afde9
import streamlit as st
import torch
import transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
@st.cache(hash_funcs=
{transformers.models.gpt2.tokenization_gpt2_fast.GPT2TokenizerFast: hash},
suppress_st_warning=True, allow_output_mutation=True)
def load_data():
tokenizer = AutoTokenizer.from_pretrained("Rubiksman1006/gpt-neo-2.7b-monika-fp16")
model = AutoModelForCausalLM.from_pretrained("Rubiksman1006/gpt-neo-2.7b-monika-fp16")
return tokenizer, model
tokenizer, model = load_data()
st.write("Welcome to the Chatbot. I am still learning, please be patient")
input = st.text_input('User:')
if 'count' not in st.session_state or st.session_state.count == 6:
st.session_state.count = 0
st.session_state.chat_history_ids = None
st.session_state.old_response = ''
else:
st.session_state.count += 1
new_user_input_ids = tokenizer.encode(input + tokenizer.eos_token, return_tensors='pt')
bot_input_ids = torch.cat([st.session_state.chat_history_ids, new_user_input_ids], dim=-1) if st.session_state.count > 1 else new_user_input_ids
st.session_state.chat_history_ids = model.generate(bot_input_ids, max_length=5000, pad_token_id=tokenizer.eos_token_id)
response = tokenizer.decode(st.session_state.chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)
if st.session_state.old_response == response:
bot_input_ids = new_user_input_ids
st.session_state.chat_history_ids = model.generate(bot_input_ids, max_length=5000, pad_token_id=tokenizer.eos_token_id)
response = tokenizer.decode(st.session_state.chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)
st.write(f"Chatbot: {response}")
st.session_state.old_response = response
|