Language translation with pretrained PyTorch model
Your team at PyBooks is working on an AI project that involves translation from one language to another. They want to leverage pre-trained models for this task, which can save a lot of training time and resources. The task for this exercise is to set up a translation model from HuggingFace's Transformers library, specifically the T5 (Text-To-Text Transfer Transformer) model, and use it to translate an English phrase to French.
T5Tokenizer
, T5ForConditionalGeneration
have been loaded for you.
This exercise is part of the course
Deep Learning for Text with PyTorch
Exercise instructions
- Initialize the
tokenizer
andmodel
from the pretrained"t5-small"
model. - Encode the input prompt using the tokenizer, making sure to return PyTorch tensors.
- Translate the input prompt using
model
and generate the translated output.
Hands-on interactive exercise
Have a go at this exercise by completing this sample code.
# Initalize tokenizer and model
tokenizer = ____.from_pretrained("t5-small")
model = ____.from_pretrained("t5-small")
input_prompt = "translate English to French: 'Hello, how are you?'"
# Encode the input prompt using the tokenizer
input_ids = ____.____(input_prompt, return_tensors="____")
# Generate the translated ouput
output = model.____(input_ids, max_length=50)
generated_text = tokenizer.decode(output[0], skip_special_tokens=True)
print("Generated text:",generated_text)