Knowledge Check 4 Question 1

Assume you are training a small language model on the following 10 sentences that vary in length.

"I love NLP" (3 words)
"Natural language processing is fun" (5 words)
"Transformers are great" (3 words)
"We learn about n-grams and padding" (6 words)
"Probability plays a key role" (5 words)
"Batching helps with computational efficiency" (5 words)
"Padding sequences ensure uniform lengths" (5 words)
"Knowing your dataset is important" (5 words)
"Stochasticity introduces randomness" (3 words)
"Language models generate predictions" (4 words)

If you pad all sequences such that they are all as long as the longest sequence, how many pad tokens do you have to insert in total across all sequences?

Bài tập này là một phần của khóa học

Google DeepMind: Build Your Own Small Language Model

Xem khóa học

Bài tập tương tác thực hành

Biến lý thuyết thành hành động với một trong các bài tập tương tác của chúng tôi

Bắt đầu bài tập

Google DeepMind: Build Your Own Small Language Model

SkillTag.level.intermediateSkillTag.label

4.7+

25 reviews

In this module, you will move beyond the manual method and explore how n-grams can be used to tokenize data. You will investigate how probabilities can be calculated to begin identifying language patterns. You will then build your own n-gram model using a small dataset and examine its limitations. Furthermore, you will consider the process researchers undertake when approaching real-world problems through the lens of Google DeepMind’s AlphaFold project. Finally, you will reflect on your own values and those of your community, as well as the role AI systems play in making decisions that involve ethical choices.

Exercise 1: N-grams Exercise 2: Lab: Experiment with N-grams Exercise 3: The limitations of n-grams Exercise 4: AlphaFold: The power of machine learning Exercise 5: Weighing values: Culture and ethics in the trolley problem Exercise 6: Applying a local ethical lens to the trolley problem Exercise 7: Knowledge Check 2 Question 1 Exercise 8: Knowledge Check 2 Question 2

In this module, you will experiment with more sophisticated transformer models and evaluate how they perform in comparison to n-gram models. You will take a deeper dive into the anatomy of language models and their core components. You will continue reflecting on the role that values play in guiding which technical problems you choose to solve. Specifically, you will consider the Ubuntu moral system and compare its characteristics with moral values popular in Europe and North America. Finally, you will design a values framework for guiding LLM development in your local community.

Exercise 1: Lab: Compare N-Gram Models and Transformer Language Models Exercise 2: Core aspects of Ubuntu Exercise 3: Develop a local values framework Exercise 4: Anatomy of a language model Exercise 5: What does it mean to train a model?Exercise 6: Knowledge Check 3 Question 1 Exercise 7: Knowledge Check 3 Question 2

In this module, you will consider the specific benefits that transformer LLMs can bring about for different sectors in your local context. You will then explore what makes a good problem statement before developing your own problem statement for a challenge around language models that you have identified in your community.

Exercise 1: Anticipating benefits Exercise 2: Challenge: Develop your problem statement Exercise 3: Knowledge Check 5 Question 1 Exercise 4: Knowledge Check 5 Question 2

In this module, you will have the opportunity to consult additional resources and further reading to investigate the topics you have covered in more detail. Finally, you will consider your next steps and how you can build on what you have learned in the course.

Exercise 1: Summary Exercise 2: Looking forward Exercise 3: Additional resources and further reading Exercise 4: Feedback Exercise 5: Glossary