Using the correct model structure
Question-answering can be either extractive or generative, each requiring a different transformer structure to process input and output correctly.
They use either:
- Encoder-only models such as
"distilbert-base-uncased-distilled-squad"
- Decoder-only models such as
"gpt2"
Use your knowledge of common models for specific tasks to select the appropriate one. pipeline is loaded, as well as text on the Mona Lisa.
Este exercício faz parte do curso
Introduction to LLMs in Python
Exercício interativo prático
Experimente este exercício completando este código de exemplo.
question = "Who painted the Mona Lisa?"
# Define the appropriate model
qa = pipeline(task="question-answering", model=____)
output = qa(question=question, context=text)
print(output['answer'])