Using AutoClasses
You’ve seen how tokenizers work and explored their role in preparing text for models. Now, let’s take it a step further by combining AutoModels and AutoTokenizers with the pipeline() function. It's a nice balance of control and convenience.
Continue with the sentiment analysis task and combine AutoClasses with the pipeline module.
AutoModelForSequenceClassification, AutoTokenizer and pipeline from the transformers library have already been imported for you.
This exercise is part of the course
Working with Hugging Face
Exercise instructions
- Download the model and tokenizer and save as
my_modelandmy_tokenizer, respectively. - Create the pipeline and save as
my_pipeline. - Predict the output using
my_pipelineand save asoutput.
Hands-on interactive exercise
Have a go at this exercise by completing this sample code.
# Download the model and tokenizer
my_model = ____.from_pretrained("distilbert-base-uncased-finetuned-sst-2-english")
my_tokenizer = ____.from_pretrained("distilbert-base-uncased-finetuned-sst-2-english")
# Create the pipeline
my_pipeline = pipeline(task="sentiment-analysis", ____=____, ____=____)
# Predict the sentiment
output = ____("This course is pretty good, I guess.")
print(f"Sentiment using AutoClasses: {output[0]['label']}")