YAML Metadata
Warning:
empty or missing yaml metadata in repo card
(https://huggingface.co/docs/hub/model-cards#model-card-metadata)
Finetuned DialoGPT model for Eng-Spa translation
DialoGPT-small model was used and finetuned on English to Spanish translations, extracted from http://storage.googleapis.com/download.tensorflow.org/data/spa-eng.zip
some examples of translations
Role | Response |
---|---|
User | please, sing me a song |
Bot | Por favor, canta una canción. |
User | I really want to go to China |
Bot | Realmente quiero ir a China. |
User | Can you do me a favor? |
Bot | ¿Me puedes hacer un favor? |
User | I don't know what you are talking about |
Bot | No sé de qué estás hablando. |
User | I don't want to go to China |
Bot | No quiero ir a China. |
Using the model
example code for trying out the model
from transformers import AutoModelWithLMHead, AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained('microsoft/DialoGPT-small')
model = AutoModelWithLMHead.from_pretrained('OscarNav/dialoGPT_translate')
# Let's traslate 5 sentences
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')
# generated a response while limiting the total chat history to 1000 tokens,
chat_history_ids = model.generate(
new_user_input_ids, max_length=1000,
pad_token_id=tokenizer.eos_token_id,
top_p=0.92, top_k = 50
)
# pretty print last ouput tokens from bot
print("DialoGPT: {}".format(tokenizer.decode(chat_history_ids[:, new_user_input_ids.shape[-1]:][0], skip_special_tokens=True)))
- Downloads last month
- 16