tosin/dialogpt_afriwoz_pidgin

DialoGPT_AfriWOZ (Pidgin)

This is a fine-tuned model of DialoGPT (small) on the AfriWOZ dataset. It is intended to be used as a conversational system in Nigeria Pidgin English language. The dataset it's trained on is limited in scope, as it covers only certain domains such as restaurants, hotel, taxi, and booking.

The perplexity achieved on the validation set is 38.52.

• Generation example from an interactive environment:

Role	Response
User	I hear say restaurant dey here.
Bot	I dey find restaurant.
User	Abeg you fit tell me which kind chop dey?
Bot	I no dey sure right now.
User	You do well. Thank you.
Bot	I dey find restauarant.

Please find the information about preprocessing, training and full details of the DialoGPT in the original DialoGPT repository

The paper for this work can be found on arXiv: https://arxiv.org/pdf/2204.08083.pdf

How to use

Now we are ready to try out how the model works as a chatting partner!

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
tokenizer = AutoTokenizer.from_pretrained("tosin/dialogpt_afriwoz_pidgin")
model = AutoModelForCausalLM.from_pretrained("tosin/dialogpt_afriwoz_pidgin")
# Let's chat for 5 lines
for step in range(5):
    # encode the new user input, add the eos_token and return a tensor in Pytorch
    new_user_input_ids = tokenizer.encode(input(">> User:") + tokenizer.eos_token, return_tensors='pt')
    # append the new user input tokens to the chat history
    bot_input_ids = torch.cat([chat_history_ids, new_user_input_ids], dim=-1) if step > 0 else new_user_input_ids
    # generated a response while limiting the total chat history to 1000 tokens, 
    chat_history_ids = model.generate(bot_input_ids, max_length=1000, pad_token_id=tokenizer.eos_token_id)
    # pretty print last ouput tokens from bot
    print("DialoGPT_pidgin_Bot: {}".format(tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)))