LLM Midterm Quiz 1

Which function can one use to typecast a tensor?

to

GPT Models are capable of translation tasks (T/F)

True

Which one of the following is not part of stage 1 of coding an LLM?

Model Evaluation

Open source models offer mort benefits than proprietary models (T/F)

True

Masked self attention prevents information leaking (T/F)

True

Tensor Library

PyTorch Implements a tensor (array) library for efficient computing

Deep learning library

PyTorch's deep learning utilities make use of its library and automatic differentiation engine

Automatic differentiation engine

PyTorch includes utilities to differentiate computations automatically

Pretrained LLMs (foundation model) are good at? (select choices that apply - could be multiple answers) Correct answer:

Text Completion

Translation

Classification

Few-shot capabilities

Personal Assistant

Summarization

Text Completion

Few-shot capabilities

Match each of the PyTorch functions below to their meanings len

return total length of the dataset

Match each of the PyTorch functions below to their meanings. init

set up attributes (file paths, file objects, database connectors, etc.,)

getitem

retrieving exactly one data record and the corresponding label

The ability to perform tasks that the model wasn’t explicitly trained to perform is called ____

Emergent Behavior

Which one of the following is not part of stage 2 of coding an LLM

Finetuning

Pretraining is generally less compute-intensive and requires less data. (T/F)

False

instruction fine-tuning

labeled dataset consists of instruction and answer pairs

few-shot

completes a task given a few examples of the task

classification fine-tuning

labeled dataset consists of texts and associated class labels

zero-shot

complete a task without an explicit example

text completion

creates plausible text given a partial input text