챗봇과 대화하기

!pip install transformers

import tensorflow as tf
from transformers import TFGPT2LMHeadModel

from google.colab import drive
drive.mount('/gdrive')

Mounted at /gdrive

chat_model = TFGPT2LMHeadModel.from_pretrained('/gdrive/My Drive/kogpt2/kogpt2_chatbot')

All model checkpoint layers were used when initializing TFGPT2LMHeadModel.

All the layers of TFGPT2LMHeadModel were initialized from the model checkpoint at /gdrive/My Drive/kogpt2/kogpt2_chatbot.
If your task is similar to the task the model of the checkpoint was trained on, you can already use TFGPT2LMHeadModel for predictions without further training.

from sentencepiece import SentencePieceProcessor

tokenizer = SentencePieceProcessor(model_file='/gdrive/My Drive/kogpt2/kogpt2_news_wiki_ko_cased_818bfa919d.spiece')

text = '오늘도 좋은 하루!'

sent = f'Q: {text} A:'

sent

'Q: 오늘도 좋은 하루! A:'

input_ids = [tokenizer.bos_id()] + tokenizer.encode_as_ids(sent)
input_ids = tf.convert_to_tensor([input_ids])

output = chat_model.generate(input_ids, max_length=50, early_stopping=True, eos_token_id=tokenizer.eos_id())

Setting `pad_token_id` to 1 (first `eos_token_id`) to generate sequence

full = tokenizer.decode_ids(output[0].numpy().tolist())

answer = full[len(sent):]

answer.strip()

'좋은 하루 되세요.'

output = chat_model.generate(input_ids, max_length=50, do_sample=True, top_k=50, early_stopping=True, eos_token_id=tokenizer.eos_id())
tokenizer.decode_ids(output[0].numpy().tolist())

Setting `pad_token_id` to 1 (first `eos_token_id`) to generate sequence

'Q: 오늘도 좋은 하루! A: 시간이 너무 빨리 지나갔네요.'