Bedrock and GenAI Flashcards

1
Q

What is Bedrock used for?

A

To create GenAI applications on AWS

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

Is Bedrock fully managed?

A

Yes

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

What is Amazon’s foundation model?

A

Amazon Titan

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

How many foundation models does Bedrock support?

A

At least seven including their own.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

When you choose a foundation model in Bedrock, is it shared?

A

No, it makes a copy of the foundation model which is only available to you which you can further fine tune with your own data.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

When using a foundation model in Bedrock, is your data ever sent back to train the model?

A

No. Your data is never used to train the model?

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

What is the playground in Bedrock?

A

A place for you to test a specific model based on your input. It helps you select your

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

Can you access Titan through an API?

A

Yes

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

What types of inputs does Amazon Titan support?

A

Image, Text, and multmodal

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

Can Amazon Titan be customized with your own data?

A

Yes

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

What are the pros and cons of smaller foundation models?

A

They are cheaper, but know less.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

What is the maximum amount of tokens for Amazon Titan?

A

8K

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

What is a continued pre-training job in Bedrock?

A

It continuously trains your model on the newest data.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

What is a fine-tuning job in Bedrock?

A

It is a one-time training job on your custom data.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

Are all foundation models able to be fine-tuned or trained on custom data?

A

No

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

Where does custom data for fine-tuning jobs for Bedrock need to be stored?

A

In S3

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
17
Q

What are hyperparameters?

A

They are configuration parameters

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
18
Q

Where is validation data for a fine-tuning job stored?

A

In S3

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
19
Q

Does fine-tuning require provisioned throughput?

A

Yes. This is expensive.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
20
Q

What does model fin-tuning do under the hood?

A

It changes the weight of the base foundation model.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
21
Q

What is instruction-based fine-tuning?

A

It improves the performance of a pre-trained model on domain specific tasks. It uses labeled examples that are prompt-response pairs.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
22
Q

What is continued pre-training?

A

This is also called domain-adaption fine-tuning. It makes a model an expert in a specific domain. E.g., ingesting all AWS documentation.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
23
Q

What is single-turn messaging?

A

Part of instruction-based fine-tuning.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
24
Q

What is multi-turn messaging?

A

Part of instruction-based fine-tuning. Good for larger context.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
25
Q

What is Transfer Learning?

A

Use a pre-trained model to adapt it to a new related task. Wiidely used for image classification and NLP.

26
Q

What is Automatic Evaluation in Amazon Bedrock?

A

It allows you to evaluate a model for quality control. You can use your own questions or AWS’

27
Q

What are benchmark datasets?

A

They are curated collections of data designed to evaluate the performance of a language model.

28
Q

What do benchmark datasets help measure?

A

accuracy, speed, efficiency, and scalability.

29
Q

Can a benchmark dataset detect bias?

A

Yes, some can.

30
Q

What is the metric ROUGE used for?

A

Evaluating automatic summarization and machine translation systems in a foundation model.

31
Q

What is the metric BLEU for?

A

Evaluates the quality of generated text, especially for translations in a foundation model.

32
Q

What is the metric BERTscore for ?

A

Semantic similarity between generated text in a foundation model.

33
Q

Where can you perform model evaluation?

A

Within Bedrock

34
Q

What is RAG?

A

Retrieval-Augmented Generation.. This allows a foundation model to reference a datasource outside of its training data.

35
Q

Where does Bedrock store RAG data that was loaded in S3?

A

Into a Knowledge Base, which is a vector database for retrieval.

36
Q

Does RAG require foundation model fine-tuning?

A

No. It is not required.

37
Q

What is a good use case for RAG?

A

When the model needs access to real-time data.

38
Q

What AWS services can be used for the vector database required for RAG?

A

OpenSearch and DocumentDB, Aurora, and RDS for PGSQL

39
Q

What is an embedding model in the context of RAG?

A

It converts the data into vectors for the vector database.

40
Q

What are some data sources for RAG?

A

S3, Confluence, Sharepoint, Salesforce, Web Pages

41
Q

What is a good use case for RAG?

A

A customer service chatbot that will have access to your products, features, and specifications.

42
Q

Can you create a knowledge base with a root user?

A

No. You must use an IAM user

43
Q

What is tokenization?

A

It turns raw text into a sequence of tokens.

44
Q

What is word-based tokenization?

A

Text is split into individual words?

45
Q

What is subword tokenization?

A

Some words can be split. This is useful for long words.

46
Q

What is a context window?

A

It is the number of tokens an LLM can consider when generating text.

47
Q

Do large context windows cause high memory and processing spikes?

A

Yes

48
Q

What is embedding?

A

It creates vectors out of text, images, or audio using embedding models to insert into a vector database.

49
Q

What are embedding models good for?

A

They can power search applications.

50
Q

What are Bedrock Guardrails?

A

They control the interactions between users and foundation models.

51
Q

Can Bedrock guardrails filter undesirable and harmful content?

A

Yes.

52
Q

Can Bedrock Guardrails remove PII?

A

Yes

53
Q

Can you create multiple Bedrock Guardrails?

A

Yes

54
Q

Can you view all the user inputs that can violate Bedrock Guardrails?

A

Yes

55
Q

Can Bedrock Guardrails mask emails?

A

Yes

56
Q

What are Bedrock Agents?

A

They manage and carry out various multi-step tasks related to infrastructure provisioning, application deployment, and operational activities.

57
Q

What are Agent Action Groups?

A

They are pre-defined action groups for agents to perform. For example, one action group can query a database for a contact name and another can be an API call to get recent orders.

58
Q

What is model invocation logging?

A

It sends logs of all invocations to the CloudWatch logs. It can include text, images, and embeddings.

59
Q

Are Bedrock metrics sent to CloudWatch?

A

Yes

60
Q

What is Bedrock Studio?

A

It gives access to your team so they can easily create AI powered applications.

61
Q

What is Batch Mode in Bedrock?

A

Allows you to make multiple predictions at a time and can provide discounts up to 50%. Takes longer for a response.

62
Q

What is the main driver of costs in Bedrock?

A

Number of input and output tokens