1
Q
Portion of training training data is labeled and feedback is provided in the form of rewards or penalties. What type of learning
A
Reinforcement learning
2
Q
What are the two types of inferencing?
A
Batch and Real time
3
Q
Deep learning is used in which use 2 cases?
A
Computer vision and NLP
4
Q
What are FMs in generative AI?
A
Pretrained models
5
Q
What are Transformer models?
A
Builds encoder decoder concept in genAI. They use self-attention to process input data. Self-attention allows the model to weigh the importance of different words in a sentence when encoding a particular word
6
Q
Are FMs pre trained using reinforced learning? True or False?
A
False. FMs are typically pre-trained through self-supervised learning
7
Q
Where is pre-text tasks used for?
A
In self-supervised learning
8
Q
Self-supervised learning makes use of the structure within the data to autogenerate labels. True or False?
A
True
9
Q
Optimization of pre trained FMs are done using what?
A
Prompt engineering,
Retrieval-augmented generation (RAG),
Fine-tuning on task-specific data
10
Q
LLMs, Diffusion and Multiodel models are what?
A
These are FM models
11
Q
These are numerical representations of tokens, where each token is assigned a vector (a list of numbers) that captures its meaning and relationships with other tokens?
A
Embeddings
12
Q
What is a context window?
A
The maximum number of tokens a LLM model can take when generating text
13
Q
What is a vector?
A
It is an array of numercial values
14
Q
What is the process of vectorization?
A
Text -> [Tokenization]->Tokens -> [Embeddings Model] -> Vectors
15
Q
What is the process of vectorization in Bedrock KBs using RAG?
A
Customer KB->[Upload in Amazon S3]->[Select a vector DB]->[Select a Model]->[Sync with customer KB]->Vectorization of Customer KB text
16
Q
What is Watermark detection for Amazon Bedrock?
A
Identify images generated by Amazon Titan Image Generator, a foundation model that allows users to create realistic, studio-quality images in large volumes and at low cost, using natural language prompts
17
Q
What is continued pretraining in Amazon Bedrock?
A
You provide unlabeled data to pre-train a model by familiarizing it with certain types of inputs
18
Q
Which is the models which gradually add more and more meaningful information to this noise until they end up with a clear and coherent output, like an image or a piece of text?
A
Diffusion model
19
Q
Which model has generator and discriminator?
A
Generative adversarial networks
20
Q
Which model has encoders and decoders?
A
Varional autoencoders
21
Q
What are the components of prompt engineering
A
Instructions, Context, Input data and Output indicator
22
Q
What are non-determistic LLMs popularly called?
A
Generative Language Models
23
Q
What is a supervised learning process that involves taking a pre-trained model and adding specific, smaller datasets?
A
Fine tuning
24
Q
Two types of fine tuning
A
Instruction fine-tuning and Reinforcement learning from human feedback (RLHF)
25
Fine tuning does it add weight to the data?
Yes
26
What is Retrieval-augmented generation (RAG)?
Supplies domain-relevant data as context to produce responses based on that data.
27
To create fine tuned models in Bedrock what is the pricing option?
Provisioned Throughput only which billed by the hour
28
How is RAG different from fine tuning
Rather than having to fine-tune an FM with a small set of labeled examples, RAG retrieves a small set of relevant documents and uses that to provide context to answer the user prompt
29
What are two types of supervised learning?
Classification and Regression
30
Predicting continuous or numerical values based on one or more input variable?
Regression
31
Forcasting uses which supervised learning technique?
Regression
32
Diagnostic uses which supervised learning technique?
Classfication
33
What are two types of unsupervised learning?
Clustering and Dimensionality reduction
34
Examples of RAG vector databases?
Amazon OpenSearch Service(KNN capability, vector embeddings),
DynamoDB(high performance,vector embeddings),
Aurora(RDS),
RDS for PostgreqSQL(RDS and open source),
Neptune(GraphQL)
35
Grouping of unstructured data is done in which type of unsupervised learning?
Clustering
36
Reducing the number of features or dimensions in a dataset in which type of unsupervised learning?
Dimensionality reduction
37
Which learning type continuously improves its model by mining feedback from previous iterations?
Reinforcement learning
38
In which learning the reward of a desired outcome is known, but the path to achieving it isn’t?
Reinforcement learning
39
How to reduce toxity risk in generative AI?
Use guardrail models
40
What does guradrail models do?
These models will detect and filter out unwanted content
41
What is the risk term for when model generates inaccurate responses that are not consistent with the training data?
Hellucinations
42
What is the risk term when model might generate different outputs for the same input?
Nondeterminism
43
What is the risk term when the information shared with your model can include personal information and can potentially violate privacy laws?
Data security and privacy concerns
44
What is the risk term when output generated by model has PII?
Regulatory violations
45
Which generative AI model used for chatbots?
Llama
46
Which generative AI model used for code generation?
Claude
47
Which generative AI model used for code gaming?
Stable Diffusion
48
Which generative AI has embeddings?
Amazon Titan
49
Which generative AI has a use case of Healthcare – summarize key ideas from long text?
Command
50
What are the capabilities of generative AI?
SPARCD
Adaptability
Responsiveness
Simplicity
Creativity and exploration
Data efficiency
Personalization
51
Recommendation engines, gaming, and voice assistance are examples of which type of AI system?
Traditional AI
52
Chatbots, code generation, and text and image generation are examples of which type of AI system?
Generative AI
53
When is a model is underfitted?
When a model has a high bias
54
Overfitting happens when?
When model performs well on the training data but does not perform well on the evaluation data
55
SageMaker Clarify
You can automatically evaluate FMs for your generative AI use case with metrics such as accuracy, robustness, and toxicity to support your responsible AI initiative
56
SageMaker Clarify is used for text based models only. True or False?
True
57
Model evaluation on Amazon Bedrock
Evaluate, compare, and select the best foundation model for your use case in just a few clicks
58
Can in human evaluation, we can automate it?
Yes, using built in task types
59
Amazon Bedrock Guardrails
Guardrails helps control the interaction between users and FMs by filtering undesirable and harmful content, redacting personally identifiable information (PII), and enhancing content safety and privacy in generative AI applications
60
Amazon SageMaker Data Wrangler
Offers three balancing operators: random undersampling, random oversampling, and Synthetic Minority Oversampling Technique (SMOTE) to rebalance data in your unbalanced datasets
61
Amazon SageMaker Experiments
Provide scores detailing which features contributed the most to your model prediction on a particular input for tabular, natural language processing (NLP), and computer vision models
You can use to create, manage, analyze, and compare your machine learning experiments.
62
Amazon A2I
Human review of ML predictions
63
SageMake governance tools
Amazon SageMaker Role Manager - define minimum permissions in minutes
Amazon SageMaker Model Cards - capture, retrieve, and share essential model information, such as intended uses, risk ratings, and training details, from conception to deployment
Amazon SageMaker Model Dashboard - You can keep your team informed on model behavior in production, all in one place
64
AWS AI Service Cards
Responsible AI documentation
Basic concepts to help customers better understand the service or service features
Intended use cases and limitations
Responsible AI design considerations
Guidance on deployment and performance optimization
65
What are the four elements of responsible agency in responsible AI?
Value alignment
Responsible reasoning skills
Appropriate level of autonomy
Transparency and accountability
66
What is Data curation?
Curating datasets is the process of labeling, organizing, and preprocessing the data
67
What are the three steps of data curation?
Data preprocessing, augmentation and audit
68
Whar are the AWS tools for transparency?
AWS AI Service Cards and Amazon SageMaker Model Cards
69
Whar are the AWS tools for explainability?
SageMaker Clarify and SageMaker Autopilot
70
What to use to build your custom ML model in SageMaker with less code?
Sagemaker Canvas using AutoML powered by Sagemaker Autopilot
71
Which is more detailed interpretability or explainability ?
Interpretability
72
PDP graphs are used for what?
They tell how single feature influence the predicted outcome. Used for interpretibiity and explainability
73
What are human centered design concepts?
Design for amplified decision making
Design for unbiased decision making
Design for human and AI learning
74
This principle seeks to maximize the benefits of using technology while minimizing potential risks and errors, especially risks and errors that can occur when humans make decisions under stress or in high-pressure environments
Design for amplified decision-making
75
The design of decision-making processes, systems, and tools is free from biases that can influence the outcomes
Design for unbiased decision-making
76
This priciple aims to create learning environments and tools that are beneficial and effective for both humans and AI
Design for human and AI learning
77
Reinforcement learning from human feedback
(RLHF) is an ML technique that uses human feedback to optimize ML models to self-learn more efficiently
78
Which AWS tool provide RLHF?
Amazon SageMaker Ground Truth
79
What is Feature engineering?
It is the process of creating, transforming, extracting, and selecting variables from data. Convert raw data into meaningful data
80
Common ratio for data training, validation and testing?
80,10,10 or 70,15,15
81
Amazon Sagemake which part does "Collecting, analyzing, and preparing your data"?
Amazon SageMaker Data Wrangler
82
Amazon Sagemake which part does "Managing Features"?
Amazon SageMaker Feature Store
83
Amazon Sagemake which part does "Model training and evaluation"
Amazon SageMaker Canvas
84
Amazon SageMaker Canvas
Access to ready models from Bedrock and Jumpstart and No coding is required
It is integrated with Comprehend, Rekognition and Textextract
85
Amazon SageMaker JumpStart?
Provides pretrained, open source models that customers can use for a wide range of problem types
86
Amazon Sagemaker which part does "Model evaluation"
Amazon SageMaker Experiments
87
Amazon Sagemaker which part does "Hyperparameter tuning"
Amazon SageMaker Automatic Model Tuning
88
What is learning rate hyperparameter?
So if you have a higher learning rate, that means that your model is going
to have a faster conversions, but there is a risk of you to overshoot the optimal solution because while you're going too fast for learning.
And if you have a low learning rate, it may be more precise but slower convergence
89
Amazon Sakemaker which part does "Monitoring"
Amazon SageMaker Model Monitor
90
Supervised learning Sagemaker built in algorithms
(FxKL)
Linera learner
Factorization machines
XGBoost
K-Nearst Neighbours(KNN)
91
Unsupervised learning Sagemaker built in algorithms
Clustering - K-means, LDA
Topic Modeling - LDA
Embeddings - Object2Vec
Anomoly detection - Random cut forest, IP insights
Dimensionality reduction - Pricipal component analysis (PCA)
92
Image processing sagemake built in algorithms
Image classification - MXNet tensor flow
Object detection - MXNet tensor flow
Semantic segmentation - FCN,PSP,Deeplab V3
Time series - DeepAR
93
Text Analysis sagemake built in algorithms
Text classification -Blazing text
Word2Vec - Blazing text
Machine translation - Sequence to sequence
Topic modeling - LDA,NTM
Speech - Sequence to sequence
94
When there is greater gap between predicted and actual value what it means?
High bias
95
When the predicted values are too much dispersed what it means?
High variance
96
How to reduce high variance?
Feature selection for more important features and multiple sets of training and test sets of data
97
What matrix help classify why and how a model gets something wrong?
Confusion matrix
98
Confusion matrix?
It is used to evaluate the performance of the model that does classfication
99
Formula for accuracy
(TP+TN)/(TP+FP+TN+FN)
100
Formula for Precision
(TP)/(TP+FP)
101
When to use precision over accuracy?
When the cost of false positives are high in your particular business situation
Think about a classification model that identifies emails as spam or not. In this case, you do not want your model labeling a legitimate email as spam and preventing your users from seeing that email.
102
Formula for Recall
(TP)/(TP+FN)
103
When to use recall over precision?
If it is extremely important and vital to the success of the model that it not give false negative results
Think about a model that needs to predict whether a patient has a terminal illness or not
104
Formula for AUC-ROC
ROC is a probability curve, and AUC represents the degree or measure of separability.
In general, AUC-ROC can show what the curve for true positive compared to false positive looks like at various thresholds.
105
What is Mean squared error?
You take the difference between the prediction and actual value, square that difference, and then sum up all the squared differences for all the observations and divide by number of predictions
106
What is R squared?
R squared explains the fraction of variance accounted for by the model
107
What is the difference between Mean squared error and R squared?
MSE focuses the measure of model performance.
R squared provides a measure of the model's goodness of fit to the data.
108
What is A/B testing or the canary deployments technique?
Developers can experiment with two or more variants of a model and help achieve the business goals.
109
For real time interactive workloads with low latency requirements, which deployment model?
Real-time
110
Requests with large payload sizes(upto 1GB), high processing time and near real time latency requirement, which deployment model?
Asynchronous
111
Large dateset and don't need a persistent endpoint, which deployment model?
Batch transform
112
Workloads that have idle periods and can tolerate cold starts, which deployment model?
Serverless
113
Providing self-service enviornments and curated data sets, which benfit of MLOps?
Productivity
114
Incoporating CI/CD practices, which benfit of MLOps?
Reliability
115
MlFlow
Manage entire ML lifecycle in Sagemaker Studio
116
Automating all steps in ML development lifecycle, which benfit of MLOps?
Repeatibility
117
Versioning all inputs and outputs , which benfit of MLOps?
Auditibility
118
Policies to guard against model bias and track changes to data statistical properties, which benfit of MLOps?
Quality
119
Sagemaker pipeline
Prepare data:
Sagemake data wrangle
Sagemake processing job
Curate feature:
Sagemake feature store
Experiment tracking:
Sagemaker experiments
Train model:
Sagemaker training job
Evaluate model:
Sagemaker processing job
Register model:
Sagemaker model registry
Deploy model:
Deployments
Manage model:
Sagemaker model monitor
120
GenerativeAI pretrained models - J2
LLM - text generation, contextual question answering, summarization, and classification
121
GenerativeAI pretrained models - Amazon Titan
Embeddings, text generation, and image generation
122
GenerativeAI pretrained model - Claude
Art vision and text AI models
123
GenerativeAI pretrained model - Command XL
Text-based responses based on prompts
124
GenerativeAI pretrained models -Llama 3
LLM - generate coherent and contextually relevant text
125
GenerativeAI pretrained models -Mistral Large
large reasoning capabilities or are highly specialized, like synthetic text generation, code generation, RAG, or agents
126
GenerativeAI pretrained models -Stable Diffusion
Can generate images of from text input.
127
How to increase the performace of the model?
Using prompt engineering, RAG, fine-tuning, or automation agents
128
Prompt engineer works for which models?
LLMs
129
What is demonstrations, or task-specific instructions called in prompt engineering?
Augmentation
130
What is Iteratively refining and adjusting the prompts called?
Refining
131
What is combining multiple prompts or generation strategies?
Ensembling
132
What is prompt searching, prompt generation, or prompt retrieval from large prompt libraries?
Mining
133
Three use cases of RAG
1. Customer support, virtual assistants
2. Journalism and research
3. Content marketing
134
What is Amazon Bedrock knowledge base?
Provide you the capability of amassing data sources into a repository of information
135
What model optmization technique used by Amazon Bedrock knowledge bases?
RAG without customization
136
Pricing of Amazon Bedrock
Pay as you go
Based on no of tokens in input and response for text based models
Pay as you go
Based on no of images in input and response for image based models
Batch
Multiple predictions at a time and sent as 1 file to S3
50% discount
Provisioned Throughput
Based on no of input and response tokens processed each minute for text based models and is called Provisioned Throughput
137
What is Fine-tuning?
Taking a pre-trained language model and further training it on a specific task or domain-specific dataset
138
What is another name of instruction based fine tuning?
Prompt Tuning
139
Instruction based find tuning features
Labeled examples
Prompt-response pairs
140
Continued Pre-training features
Particular field or area of knowledge
Unlabeled data
Domain specific training
141
What is another name of human feedback data used in fine tuning?
Reinforcement learning from human feedback (RLHF)
142
Which type of learning creates a reward model?
RLHF
143
Steps in creating a model from scratch?
1. Selecting the appropriate neural network architecture, layers, and hyperparameters
2. A large and diverse dataset must be curated, cleaned, and preprocessed
3. Model is initialized with random weights and trained using various optimization algorithms
144
What are agents used during performance improvement of a model?
Carry out various multi-step tasks related to infrastructure provisioning, application deployment, and operational activities.
Task coordination
Reporting and logging
Scalability and concurrency
Integration and communication
145
Model improvement techniques and pricing impact
Cheapest to Costly
Prompt Engineering - No model training needed
RAG - Use external knowledge but no FM changes. Cost for using vector dbs
Instruction based fine tuning - FM is fine tuned with instructions and change the tone of the model.
Domain Adaption fine tuning - Domain specific model training
146
Does changing temperature, Top K, Top P affects model pricing?
No
147
Does temperature, Top K, Top P affects model latency?
No
148
Types of benchmark datasets for generativeAI evaluation?
GLUE - text classification, question answering, and natural language inference
SuperGLUE - compositional language understanding
SQuAD - question-answering capabilities
WMT - machine translation systems
149
What are three types of automated metrics?
Perplexity (a measure of how well the model predicts the next token)
BLEU score (for evaluating machine translation)
F1 score (for evaluating classification or entity recognition tasks)
150
Formula for F1?
(2xPrecisionxRecall) / (Precision+Recall)
151
What is the advantage of automated metrics?
Automated metrics can be useful for rapid iterations and fine-tuning during model development
152
What are the disadvatage of automated metrics?
They fail to capture the nuances and complexities of human language and might not align perfectly with human judgments
153
ROUGE automated metrics definition?
Automatic summarization and machine translation systems
The main idea behind ROUGE is to count the number of overlapping units
154
BLEU automated metrics definition?
Similarity between a generated text and one or more reference translations
Used to evaluate the quality of text that has been machine-translated from one natural language to another
155
BERTScore automated metrics definition?
Compute contextualized embeddings for the input texts, and then calculates the cosine similarity between them
It relies on semantic similarity rather than relying on exact lexical matches
156
ROUGE Vs BLEU Vs BERTScore Vs Perplexity
Compare N-gram matches
Vs
Evaluate Quality(Prcesion and penalizes)
Vs
Semantic similarity(Compare embeddings)
Vs
How confident the model to predict next token(lower is better)
157
What is automated metrics used for?
Assess the performace of a FM in text summarization, machine translation, and open-ended text generation
158
What is negative prompting?
Negative prompting is used to guide the model away from producing certain types of content or exhibiting specific behaviors
159
What are the key elements in a prompt?
Intructions, Context, Input data and desired output
160
What is used to influence the model response?
Inference parameters
161
Randomness and diversity parameters?
Temperature - A higher temperature makes the output more diverse and unpredictable, and a lower temperature makes it more focused and predictable
Top P - With a low top p setting, like 0.250, the model will only consider words that make up the top 25 percent of the total probability distribution. Higher P means more diverse
Top K - Set to 50, the model will only consider the 50 most likely words for the next word in the sequence
162
Length and stop sequence in inference parameter?
Maximum length - Used in text summarization and translation
Stop Sequence - When the model encounters a stop sequence during the inference process, it will terminate the generation regardless of the maximum length setting
163
Inferencing at the edge?
Running Small Language Models on an edge device
164
Zero-shot, few shots and CoT prompting?
Zero-shot - Present task to generative model w/o and example
Few-shot - Present task to generative model with some examples
Chain of Thought - Divides intricate reasoning tasks into smaller, intermediary steps
165
Prompt misuse - Poisoning, hijacking, and prompt injection
Poisoning - intentional introduction of malicious or biased data
Hijacking, and prompt injection - influencing the outputs of generative models by embedding specific instructions
166
Prompt misuse - Exposure
Risk of exposing sensitive or confidential information from its training corpus
167
Prompt misuse - Prompt leaking
Exposing the prompt or inputs used within the model or data used by the model
168
Prompt misuse - Jailbreaking
Modifying or circumventing the constraints and safety measures implemented in a generative model or AI assistant to gain unauthorized access or functionality
169
The most common algorithms used to perform the similarity search are?
k-NN or cosine similarity
170
Name a few vector databases offered by Amazon?
Amazon Opesearch
pgvector extension in RDS
Amazon Kendra
171
What is measured by benchmark databases in model evaluation?
Accuracy
Speed and efficiency
Scalability
172
Is it true - Creating a benchmark dataset is a manual process?
True. It is a set of questions and answers provided by the SME. Model's response to the same questions is compared with benchmark datasets answers and model performance is scored
173
What type of fine tuning is this - retraining the model on a new dataset that consists of prompts followed by the desired outputs?
Instruction tuning
174
What type of fine tuning is this - model is refined through a reinforcement learning process, where a reward model built from human feedback guides the model?
Reinforcement learning from human feedback (RLHF)
175
What type of fine tuning is this - the approach involves extending the training phase of a pre-trained model by continuously feeding it new and emerging data?
Continuous pretraining
176
What are the steps in fine tuning?
Data curation
Labeling
Governance and compliance
Representativeness and bias checking
Feedback integration
177
What are the two types of ROUGE metrics
ROUGE-N - This metric primarily assesses the fluency of the text and the extent to which it includes key ideas from the reference. Compare N-gram matches between required vs actual output
ROUGE-L - It is good at evaluating the coherence and order of the narrative in the outputs. Compare the longest sequence of words matche between required vs actual output
178
What does BLEU do?
Measures the precision of N-grams in the machine-generated text that appears in the reference texts and applies a penalty for overly short translations (brevity penalty)
179
Services that provide infrastructure protection in AWS?
iAM and NACLs
180
Advantages of AI governance
Managing, optimizing, and scaling the organizational AI initiative
Maintaining responsible and trustworthy AI practices
Establish clear policies, guidelines, and oversight mechanisms
181
Proximity of data to the compute resources used for training and inference, which AI data governance data management concepts?
Data residency
182
Model performance metrics and System events, which AI data governance data management concepts?
Data logging
183
Data visualization and Exploratory data analysis (EDA), which AI data governance data management concepts?
Data analysis
184
Performance metrics used for monitoring under AI governance
Accuracy
Precision
Recall
F1-score
Latency
185
Fine tuned models vs Self trained models
Fine tuning a model using your data vs training a model from scratch using your data
186
What is source citation in GenerativeAI?
It refers to the act of properly attributing and acknowledging the sources of the data used to train the model.
Datasets
Databases
Other sources
187
What is Documenting data origins in GenerativeAI?
It provides detailed information about the provenance, or the place of origin of the data used to train the model.
Details about the data collection process
The methods used to curate and clean the data
Any preprocessing or transformations applied to the data
188
What is the technique used to track the history of data, including its origin, transformation, and movement through different systems?
Data lineage
189
What is the systematic organization and documentation of the datasets, models, and other resources used in the development of a generative AI system?
Cataloging
190
What is called the standardized format for documenting the key details about an ML model, including its intended use, performance characteristics, and potential limitations?
Model cards
191
GenerativeAI scoping matrix(From low to high ownership)
Scope 1 : Consumer App (ChatGpt)
Scope 2 : Enterprise App (SaaS like Amazon Q developer)
Scope 3: Pre-trained models (Amazon Bedrock)
Scope 4: Fine-tuned models (Amazon Bedrock customized or SageMaker Jumpstart)
Scope 5: Self trained models (SageMaker )
192
What is data engineering life cycle
Automation and access control - AWS Glue
Data collection - Kinesis, DMS, Glue
Data prep and cleaning - EMR or Glue
Data quality check - Glue data brew or Glue data quality check
Data visualization and analysis - Quicksight or Neptune
IaC deployment - CloudFormation
Monitoring and Debugging - CloudWatch
193
4 ways to build your AI solutions by chosing FM?
Reuse
Adapt
Customize
Start from scratch
194
Reuse and Adapt only in which AWS GenAI tool?
Amazon Q
195
Amazon Q pricing
Subscription based - Lite and Pro + Data storage for client documents
196
Amazon Q Admin controls?
Same as Guardrails
197
Reuse, Adapt and Customize only in which AWS AL/ML service?
Amazon Bedrock
198
Reuse, Adapt, Customize and Start from scratch
Amazon SageMaker
199
Foundation Models comparison
Amazon Titan Text Express, LLAMA 2, Claude, stability.ai
Claude can take maximum tokens - 200K
stability.ai is for image
Content creation is by Titan
Text generation and customer service by LLAMA 2
Analysis and Forecasting by Claude
Image creation by stability.ai
200
What is PartyRock
Its a playground on Amazon Bedrock to build GenAI apps
You can access without having AWS account
201
Which model used in classification?
KNN
202
Custom classfication and custom entity recognition in which AWS AI managed service?
Amazon Comprehend
203
Redaction is used in which AWS AI managed service?
Amazon Transcribe to remove PII information
204
Custome vocabularies and custome language models are used in which AWS AI managed service?
Amazon Transcribe to transcribe technical terms and jargons and context
205
Lexicons, SSML are used in which AWS AI managed service?
Amazon Polly to read specific type of text and add break, whisper etc
206
What are SLOTs use in which AWS AI managed service?
Amazon Lex to provide input parameters
207
Amazon Lex can integrate with what AWS services?
Comprehend, Connect, Lambda function & Kendra
208
AWS Questions
flashcards
Decks in class (9)
# Cards
