MLA-C01 ML Model Development — 10 Questions

Question 1

A team is fine-tuning a large language model using reinforcement learning from human feedback (RLHF) in SageMaker. Which THREE components are essential for the RLHF pipeline? (Select THREE.)

Accepted Answer

Policy network (the LLM being fine-tuned). The standard RLHF pipeline for fine-tuning an LLM requires three components: a policy network (the LLM being fine-tuned), a reward model (provides human feedback), and a reference model (used for KL divergence to prevent reward hacking). Among the given options, only the Policy network (A) and Reward model (E) are essential. The Value network (B) is not a standard component in RLHF; it is used in some reinforcement learning algorithms like PPO but not as a core part of RLHF for LLMs. The Feature store (C) and Hyperparameter tuner (D) are useful but not essential to the RLHF pipeline. Since only two of the three essential components are listed, the correct selections are A and E.

Answer

Value network

Answer

Feature store

Answer

Hyperparameter tuner

Question 2

A company is using SageMaker to train a model with a custom container. The training script requires a specific version of a Python library that is not included in the default SageMaker containers. How should they provide this library?

Accepted Answer

Extend a SageMaker framework container and install the library using a Dockerfile. Using a custom container (BYOC) allows bundling all dependencies, including specific library versions, into a Docker image that SageMaker can run.

Answer

Use SageMaker Script Mode and specify the library in a requirements.txt

Answer

Use SageMaker's lifecycle configuration to install the library on the training instance

Answer

Use pip install in the training script before model training

Question 3

A machine learning engineer wants to use SageMaker Clarify to analyze bias in their training data and model predictions. They want to detect bias before training. Which TWO types of analysis can SageMaker Clarify perform on the data?

Accepted Answer

Pre-training bias metrics (e.g., class imbalance, feature skew). SageMaker Clarify can compute pre-training bias metrics such as class imbalance and feature skew directly on the training data before any model is trained. This is the only type of analysis among the options that is performed before training. Feature importance (SHAP values) and explainability reports require a trained model, so they are post-training analyses. Model monitoring for data drift is a separate feature of SageMaker. Post-training bias metrics also require model predictions. Therefore, only option A qualifies as a pre-training analysis.

Answer

Feature importance (SHAP values) on the dataset

Answer

Model monitoring for data drift

Answer

Explainability report for the model

Answer

Post-training bias metrics (e.g., difference in accuracy across groups)

Question 4

A machine learning engineer is deploying a custom PyTorch model using SageMaker script mode. The training script requires specific dependencies not included in the default PyTorch container. Which TWO actions can the engineer take to ensure the dependencies are available? (Select TWO.)

Accepted Answer

Build a custom container that extends the SageMaker PyTorch container and push it to Amazon ECR. Building a custom container that extends the SageMaker PyTorch container and pushing it to Amazon ECR allows you to include any dependencies not available in the default container. Option B is correct: Including a requirements.txt file in the source directory causes SageMaker to automatically install those dependencies during training. Option C is incorrect: Lifecycle configurations are only applicable to notebook instances, not training jobs. Option D is incorrect: While you can specify a custom Docker image using the image_uri parameter, the action of building the container is covered by option A; option D is not a separate valid action to ensure dependencies. Option E is incorrect: The source_dir argument is for pointing to a directory containing training code, not for specifying a separate container.

Answer

Use a lifecycle configuration to install dependencies

Answer

Specify a custom Docker image in the PyTorch estimator

Answer

Add the dependencies to the estimator's source_dir argument as a separate container

Question 5

A data science team is using SageMaker Experiments to track hyperparameters and metrics for a model training project. They need to compare multiple trials and identify the best model. Which THREE actions are part of a typical workflow? (Select THREE.)

Accepted Answer

Log hyperparameters and metrics using the SageMaker SDK. Logging hyperparameters and metrics during training is essential for tracking. Option C is correct: using the SageMaker SDK to list and compare trials allows identifying the best model. Option D is correct: creating an experiment is the first step in organizing trials. Option B is incorrect: confusion matrices are not automatically generated; they must be computed manually. Option E is incorrect: deployment to an endpoint is not part of SageMaker Experiments; it is a separate step.

Answer

Generate confusion matrices for each trial automatically

Answer

Automatically deploy the best trial to an endpoint

Question 6

A company is using SageMaker built-in XGBoost algorithm for a multiclass classification problem. They want to evaluate the model's performance. Which TWO metrics are appropriate for multiclass classification? (Select TWO.)

Accepted Answer

Precision (macro or weighted). For multiclass classification problems, appropriate metrics include Precision (macro or weighted) and Recall (macro or weighted). Option B (Precision) and Option C (Recall) are correct because they can be averaged across classes to evaluate overall performance. Option A (RMSE) is for regression tasks, Option D (NDCG) is for ranking problems, and Option E (AUC) is typically used for binary classification. Therefore, only B and C are suitable.

Answer

RMSE

Answer

NDCG (Normalized Discounted Cumulative Gain)

Answer

AUC (Area Under the ROC Curve)

Question 7

A company wants to build a customer service chatbot that answers questions about their internal policy documents. The documents are updated monthly, and the team cannot afford to retrain a model each time. Which approach is MOST appropriate?

Accepted Answer

Use Retrieval-Augmented Generation (RAG) with the policy documents indexed in a vector store. RAG allows the LLM to retrieve relevant document sections at inference time, so knowledge stays current without retraining.

Answer

Use a larger foundation model with a longer context window and paste all documents into each prompt

Answer

Fine-tune a base LLM on the policy documents monthly

Answer

Train a custom model from scratch on the policy documents each month

Question 8

A data scientist is using SageMaker built-in XGBoost algorithm for a binary classification task. Which objective metric is MOST appropriate for SageMaker Automatic Model Tuning to maximize?

Accepted Answer

validation:auc

Answer

validation:mae

Answer

validation:rmse

Answer

validation:ndcg

Question 9

A team is training a large language model using SageMaker with multiple GPUs. They need to reduce training time by splitting the model across devices due to memory constraints. Which distributed training strategy should they use?

Accepted Answer

Model parallelism

Answer

SageMaker Distributed Data Parallel (SMDDP)

Answer

Data parallelism

Answer

SageMaker Distributed Model Parallel (SMDMP)

Question 10

A machine learning engineer is using SageMaker Debugger to monitor training jobs. They want to capture tensors every 100 steps but only for the first 500 steps. Which configuration should they set in the Debugger hook?

Accepted Answer

collection_configs with save_interval=100 and end_step=500

Answer

collection_configs with save_interval=500 and end_step=100

Answer

collection_configs with start_step=100 and end_step=500

Answer

Use SageMaker Debugger rules to filter steps