Applied Accelerated Artificial Intelligence - Week 0 Quiz Solutions | SwayamSolver

Back to Course: Applied Accelerated Artificial Intelligence

Practice: Week 0 : Assignment 0

Applied Accelerated Artificial Intelligence · Instructor: Prof. Satyajit Das, Prof. Satyadhyan Chickerur

Week 010 Questions

Question 1

Which of the following statement about Tensor is true: Tensor is an imperative n-dimensional array which runs on GPU. A tensor is an n-dimensional data container Tensor specifically refers to data structure part of framework Tensorflow to store arrays

A.1 and 2

B.1, 2 and 3

C.1 and 3

D.3

Correct Answer: 1 and 2

Question 2

Deep Learning Training is done in 64 bit precision

A.True

B.False

C.Never

D.Depends

Correct Answer: Depends

Question 3

Choose the correct statement:

A.Increasing Dataset for DL Training always results into better accuracy

B.It is observed empirically that increasing dataset also demands increase in model size to capture complex patterns

C.Less training dataset is key reason for Model Underfitting

D.Unsupervised learning requires a large amount of data for training

Correct Answer: It is observed empirically that increasing dataset also demands increase in model size to capture complex patterns

Question 4

BERT Base Model has how many trainable model parameters?

A.Few Million

B.Few Billion

C.Few Trillion

D.Few thousands

Correct Answer: Few Million

Question 5

Which of the following frameworks supports GPU acceleration?

A.PyTorch

B.TensorFlow

C.Matlab

D.All of the above

Correct Answer: All of the above

Question 6

Which of this is TRUE for Kubernetes

A.Open-source container orchestration tool

B.Automate tasks such as the management, monitoring, scaling, and deployment of non containerized applications

C.Kubernetes is just a a scheduler to helps launch jobs with containers

D.Kubernetes is primarily used for DL deployment

Correct Answer: Open-source container orchestration tool

Question 7

Which of this is a FALSE statement

A.GPUs provide high bandwidth RAM which helps in getter better throughput for DL training

B.GPU architecture is fundamentally parallel computing architecure and one of its usage is to accelerate AI workloads

C.GPU can be used to accelerate only DL Training and not inferencing

D.All primary DL and Data Science framework support GPU accleration and Data Scientist are not expected to learn CUDA programming

Correct Answer: GPU can be used to accelerate only DL Training and not inferencing

Question 8

Which of these is a GPU accelerated framework

A.skLearn

B.DASK

C.NumPy

D.Pandas

Correct Answer: DASK

Question 9

Which parallelism is best for scaling across multiple processors when the model is too large to fit in memory?

A.Data Parallelism

B.Model Parallelism

C.Asynchronous Model Parallelism

D.Asynchronous Data Parallelism

Correct Answer: Asynchronous Model Parallelism

Q10

Question 10

ONNX allows:

A.to help optimize the inference of your machine learning model

B.models to run on a variety of platforms and devices

C.support from different kinds of platforms (cloud/edge, CPU/GPU, etc)

D.All of the above

Correct Answer: All of the above