QuiztudyPREMIUM
🌍 All Study Guides📰 Blog💡 About
AWS Certified AI Practitioner • STUDY MODE

Practice Quiz

QUESTION 1 OF 7

A machine learning practitioner wants to transfer knowledge from a larger, highly accurate foundation model to a smaller, faster, and more cost-efficient model using Amazon Bedrock. Which customization process should they use, and how does it function?

A
Supervised fine-tuning, by adjusting the larger model's parameters using a dataset of labeled input-output pairs.
B
Distillation, which fine-tunes the smaller student model using synthesized responses generated by Amazon Bedrock from the larger teacher model.Correct Answer
C
Reinforcement fine-tuning, by utilizing AWS Lambda to evaluate reward scores on the smaller model's outputs.
D
Model storage optimization, by reducing the monthly charge of the larger model based on the number of tokens processed.
Explanation:

Distillation is a customization process in Amazon Bedrock that transfers knowledge from a larger teacher model to a smaller, faster, and more cost-efficient student model by fine-tuning the student model with synthesized responses generated by Amazon Bedrock from the teacher model. Supervised fine-tuning adjusts parameters based on labeled prompt-response pairs directly, while reinforcement fine-tuning uses reward functions to align models.

QUESTION 2 OF 7

A development team plans to customize an Amazon Bedrock foundation model for a specific business use case. To estimate their expenses, they need to know how Amazon Bedrock calculates the charges for model training and model storage. According to Amazon Bedrock pricing, how are these charges determined?

A
Model training is charged based on the number of training tokens processed (calculated as training corpus tokens multiplied by epochs), and model storage is charged monthly per model.Correct Answer
B
Model training is charged based on the AWS Lambda execution duration for reward functions, and model storage is included at no additional cost.
C
Model training is charged as a flat rate per customized model, and model storage is billed based on the number of prompts in the training prompt dataset.
D
Model training is charged per epoch regardless of the number of tokens in the corpus, and model storage is billed based on the student model size.
Explanation:

In Amazon Bedrock, charges for model training are calculated based on the number of tokens processed (training corpus tokens multiplied by the number of epochs), while model storage is charged monthly per model.

QUESTION 3 OF 7

A financial technology company wants to improve the alignment of an Amazon Bedrock foundation model. Instead of relying on a labeled dataset of input-output pairs, they want to evaluate response quality using custom reward functions. Which model customization method fits this scenario, and what inputs can they provide?

A
Supervised fine-tuning, where the company provides a training dataset of labeled examples and adjusts parameters based on AWS Lambda triggers.
B
Reinforcement fine-tuning, where the company defines reward functions using AWS Lambda and provides training prompt datasets or existing Bedrock invocation logs.Correct Answer
C
Distillation, where the company select a teacher model and a student model to synthesize labeled training prompt-response pairs automatically.
D
Model customization, where the company provides a static training corpus and is charged based strictly on the number of evaluation queries.
Explanation:

Reinforcement fine-tuning aligns foundation models via feedback-based learning using reward functions defined in AWS Lambda instead of labeled prompt-response pairs. Users provide training prompt datasets or Bedrock invocation logs.

QUESTION 4 OF 7

A machine learning team is tasked with deploying a cost-effective, fast, and lightweight generative AI model in Amazon Bedrock. To achieve this, they want to transfer knowledge from a highly intelligent, larger model to a smaller student model by automatically generating high-quality responses from the larger model and using them for fine-tuning. Which Amazon Bedrock customization method should they choose?

A
Model distillationCorrect Answer
B
Supervised fine-tuning
C
Reinforcement fine-tuning
D
Unsupervised pre-training
Explanation:

Amazon Bedrock automates the model distillation process to transfer knowledge from a larger, more intelligent teacher model to a smaller, faster, and more cost-efficient student model by generating responses from the teacher and using them to fine-tune the student.

QUESTION 5 OF 7

Under the pricing model for customizing foundation models in Amazon Bedrock, how are the charges calculated for the model training process?

A
Based on the number of processed tokens (calculated as tokens in the training corpus multiplied by the number of epochs) and monthly storage costs per model.Correct Answer
B
Based on the elapsed training time measured in hours of active GPU utilization.
C
Based on the size of the training dataset in gigabytes and the total number of model parameters modified.
D
Based on the number of invocation API requests made during the training phase.
Explanation:

Amazon Bedrock model customization charges are based on the number of processed tokens, which is calculated as the tokens in the training corpus multiplied by the number of epochs, along with model storage charged per month per model.

QUESTION 6 OF 7

A company wants to customize an Amazon Bedrock model to improve its task performance. Rather than assembling a dataset of labeled input-output examples, the company wants to use feedback-based learning where responses are evaluated dynamically. Which customization approach and AWS service integration should they use?

A
Reinforcement fine-tuning using AWS Lambda to define the evaluation reward functions.Correct Answer
B
Supervised fine-tuning using AWS Lambda to automatically label training examples.
C
Model distillation using AWS Lambda to transfer knowledge from a teacher model.
D
Continuous pre-training using AWS Lambda to stream unlabelled training data.
Explanation:

Reinforcement fine-tuning in Amazon Bedrock relies on feedback-based learning and reward functions (which can be defined using AWS Lambda) instead of using labeled input-output pairs.

QUESTION 7 OF 7

Which of the following describes the mechanism used by Amazon Bedrock for supervised fine-tuning to improve model performance?

A
It adjusts model parameters using a training dataset of labeled examples to help the model learn the association between inputs and expected outputs.Correct Answer
B
It relies on feedback-based learning and reward functions defined in AWS Lambda instead of utilizing input-output pairs.
C
It transfers knowledge from a larger teacher model to a smaller student model by generating synthetic responses.
D
It adjusts model storage configurations monthly to optimize inference performance without altering model parameters.
Explanation:

Supervised fine-tuning in Amazon Bedrock adjusts model parameters by utilizing a training dataset of labeled examples to help the model learn the association between inputs and expected outputs.

Ready to test your recall?

A machine learning practitioner wants to transfer knowledge from a larger, highly accurate foundation model to a smaller, faster, and more cost-efficient model using Amazon Bedrock. Which customization process should they use, and how does it function?

A
Supervised fine-tuning, by adjusting the larger model's parameters using a dataset of labeled input-output pairs.
B
Distillation, which fine-tunes the smaller student model using synthesized responses generated by Amazon Bedrock from the larger teacher model.
C
Reinforcement fine-tuning, by utilizing AWS Lambda to evaluate reward scores on the smaller model's outputs.
D
Model storage optimization, by reducing the monthly charge of the larger model based on the number of tokens processed.

How confident are you in this answer?