NVIDIA Generative AI Prompt Engineering Quiz

1. In transformer architectures, what is the role of the 'softmax' function?

To convert attention scores into probability distributions

To reduce model parameters

To increase training speed

To normalize input data

2. What is the primary benefit of using NVIDIA's Triton Inference Server?

Training models faster on GPUs

Deploying and managing multiple models efficiently in production

Automatically generating training data

Simplifying prompt writing for end users

3. In generative AI, what is 'beam search' used for?

Exploring multiple possible output sequences to find higher-quality results

Splitting data across multiple GPUs

Compressing model weights

Measuring inference latency

4. What role does 'gradient accumulation' play in training large generative models?

Simulating larger batch sizes when GPU memory is limited

Automatically tuning learning rates

Reducing model parameter count

Improving output quality without additional training

5. Which of the following best describes 'model hallucination'?

When a model generates plausible-sounding but false or irrelevant information

When GPU overheating causes processing errors

When training data contains duplicates

When inference speed exceeds expected performance

6. What is the purpose of 'few-shot prompting' in generative AI?

Providing a few examples in the prompt to guide the model's behavior

Limiting the number of model parameters

Reducing GPU memory consumption

Accelerating data preprocessing

7. In the context of generative AI, what does 'context window' refer to?

The maximum number of tokens a model can process in a single input

The time it takes to generate output

The number of GPU cores used

The percentage of data used for validation

8. What is the primary advantage of using NVIDIA GPUs for training generative models?

Lower cost compared to cloud services

Massive parallel processing power enabling faster training

Automatic model optimization without tuning

Built-in natural language understanding

9. Which technique helps prevent 'hallucinations' in generative AI outputs?

Increasing model size indefinitely

Using fact-checking, retrieval augmentation, and careful prompting

Reducing the number of training epochs

Removing attention mechanisms

10. What is 'retrieval-augmented generation' (RAG) used for?

Combining retrieved documents with generative models to improve response accuracy

Reducing the number of model parameters

Speeding up GPU memory access

Automating code generation

11. What is the primary purpose of prompt engineering in generative AI?

To write code for AI models

To structure input text to guide model output quality and relevance

To reduce the size of neural networks

To replace human creativity with machine generation

12. What is the purpose of a 'system prompt' in generative AI applications?

To define the model's behavior, tone, and constraints for the entire conversation

To monitor GPU temperature during inference

To automatically correct spelling errors

To limit the number of output tokens

13. Which NVIDIA tool is designed for optimizing inference performance of neural networks?

Triton Inference Server

RAPIDS

Optuna

Weights & Biases

14. What is quantization in the context of deploying generative AI models?

Reducing precision of model weights to decrease size and improve inference speed

Measuring the quality of generated text

Counting the number of training samples

Dividing data into training and validation sets

15. In prompt engineering, what is 'chain-of-thought' prompting?

Asking the model to explain reasoning step-by-step before answering

Training models on a sequence of datasets

Connecting multiple models in series

Using only short, single-sentence prompts

16. What does 'attention mechanism' enable in transformer-based generative models?

Models to focus on relevant parts of input when generating output

Faster data loading during training

Reduction in model size

Automatic hyperparameter tuning

17. Which of the following is a key ethical consideration when deploying generative AI?

Minimizing GPU memory usage

Mitigating bias and ensuring fair, responsible outputs

Maximizing inference speed

Reducing model parameter count

18. What is fine-tuning in the context of pre-trained generative models?

Adjusting model weights on task-specific data after pre-training

Removing unnecessary layers from a neural network

Converting a model to run on edge devices

Increasing the batch size during training

19. In the context of generative AI, what does 'temperature' control in model sampling?

The physical heat of GPU processors

The randomness and diversity of generated outputs

The training duration of the model

The number of parameters in the model

20. Which NVIDIA framework is primarily used for building and training large language models?

CUDA Toolkit

Megatron-LM

TensorRT

CuDNN