Finetuned DialoGPT model on Spanish Conversations

This model was finetuned from the original DialoGPT-medium model on subtitles from Spanish movies and telenovelas from the awesome OpenSubtitle dataset.

DialoGPT paper: https://arxiv.org/abs/1911.00536

Some example dialog from the model:

Role Response
User ¿Cómo está el clima hoy?
Bot Muy bien.
User Eso es bueno escuchar. ¿Vas a ir a la playa?
Bot No, también puedo regresar.
User Oh bien, bueno, iré a la playa más tarde.
Bot ¡No!
User ¿Qué está mal con eso?
Bot ¡Cómo que no!
User Estoy confundido, ¿por qué no puedo ir a la playa?
Bot ¡Cómo que no!
User Explícamelo por favor.
Bot ¡No!

Using the model

Example code for trying out the model (taken directly from the DialoGPT-medium model card):

from transformers import AutoModelWithLMHead, AutoTokenizer
import torch


tokenizer = AutoTokenizer.from_pretrained("ncoop57/DiGPTame-medium")
model = AutoModelWithLMHead.from_pretrained("ncoop57/DiGPTame-medium")

# 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: {}".format(tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)))

Training your own model

If you would like to finetune your own model or finetune this Spanish model, please checkout my blog post on that exact topic! https://nathancooper.io/i-am-a-nerd/chatbot/deep-learning/gpt2/2020/05/12/chatbot-part-1.html

Downloads last month
19
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Space using ncoop57/DiGPTame-medium 1