NVIDIA-Certified-Professional Accelerated Data Science

QUESTION NO: 1
You are running a data science project on a cloud environment, where you need to optimize the GPU utilization for real-time data processing tasks.
Which of the following practices should you consider to maximize GPU performance? (Select two)

A. Use NVIDIA CUDA libraries optimized for data science tasks to accelerate computations. B. Implement model inference in a batch mode rather than in a real-time streaming mode for better performance. C. Prioritize using CPU for computationally expensive tasks and reserve GPU for I/O operations. D. Use data parallelism techniques to split the workload evenly across multiple GPUs. E. Scale the cloud infrastructure vertically by increasing the size of the GPU in use.

Correct Answer: A,D

QUESTION NO: 2
You are tasked with profiling a deep learning model using NVIDIA's DLProf to identify performance bottlenecks and optimize resource utilization.
Which of the following statements correctly describes the capabilities of DLProf?

A. DLProf is primarily designed for debugging model accuracy rather than performance analysis. B. DLProf requires significant modifications to the source code to collect profiling data. C. DLProf only works with TensorFlow models and does not support PyTorch-based workloads. D. DLProf can generate detailed reports that highlight kernel-level execution times and GPU utilization trends.

Correct Answer: D

QUESTION NO: 3
A data scientist is working with an imbalanced dataset in a fraud detection project. The dataset contains 1 million transactions, but only 2% of them are labeled as fraudulent. To improve the performance of the model, the scientist decides to generate synthetic data using NVIDIA RAPIDS cuDF.
Which of the following approaches is the best way to generate synthetic samples while preserving data characteristics?

A. Use cudf.DataFrame.sample(frac=0.5, replace=True) to oversample the minority class. B. Use cudf.DataFrame.append(cudf.DataFrame.random()) to create new fraudulent transactions. C. Apply cuML.SMOTE() to generate synthetic samples based on the minority class distribution. D. Use cudf.DataFrame.interpolate(method='linear') to create new fraudulent samples by interpolating between existing ones.

Correct Answer: C

QUESTION NO: 4
A data scientist is working with large-scale ETL (Extract, Transform, Load) pipelines on GPU- accelerated infrastructure using RAPIDS. The workload involves frequent shuffle operations, which significantly impact performance.
What is the best approach using NVIDIA technologies to reduce shuffle overhead and improve performance?

A. Use RAPIDS cuDF's GPU memory caching to store intermediate DataFrames and avoid redundant shuffle operations. B. Store all intermediate shuffle data in CPU memory using pandas to ensure persistence and reduce GPU load. C. Use RAPIDS cuML to replace shuffle-intensive operations with an ML model that predicts data distribution. D. Enable CUDA Unified Memory to automatically optimize shuffle performance without manual intervention.

Correct Answer: A

QUESTION NO: 5
Your data science team is performing exploratory data analysis (EDA) on a large GPU-accelerated environment using cuDF and Dask-cuDF. During analysis, queries on categorical columns are performing poorly.
Which approach will most effectively improve query performance for categorical data in GPU-accelerated DataFrames?

A. Convert categorical columns to integer codes B. Increase GPU memory allocation C. Store data in compressed file formats like Parquet D. Use CPU-based DataFrames (pandas) instead of GPU-based DataFrames

Correct Answer: A

QUESTION NO: 6
You are tasked with designing and implementing a benchmark to compare the performance of different deep learning frameworks, including TensorFlow, PyTorch, and JAX, using NVIDIA GPUs.
Which of the following is the most effective approach to ensure an accurate and fair comparison?

A. Use mixed precision (FP16) training only in TensorFlow to maximize performance while keeping other frameworks at FP32. B. Ensure identical hardware configurations, dataset preprocessing, and model architectures while leveraging NVIDIA's Nsight Systems and DLProf for analysis. C. Compare training times only without considering throughput, power efficiency, or memory utilization. D. Run each framework with default settings to compare their out-of-the-box performance without any optimizations.

Correct Answer: B

QUESTION NO: 7
You are using NVIDIA DLProf to analyze the performance of a deep learning model deployed on an A100 GPU. The report indicates that compute-bound operations are dominating execution time, and kernel execution efficiency is below 50%.
What is the best action to take based on this insight?

A. Use DLProf's Tensor Core Analysis to check if the model is leveraging Tensor Cores effectively. B. Enable mixed precision training to improve computational efficiency. C. Reduce the number of layers in the model to decrease computation time. D. Increase the batch size to fully utilize available GPU memory and reduce per-sample processing overhead.

Correct Answer: A

QUESTION NO: 8
You are working on a large-scale machine learning pipeline that involves processing massive datasets using multiple GPUs on an NVIDIA DGX system. You choose to use Dask to enable efficient parallel processing across multiple GPUs.
Which of the following steps is essential to correctly configure Dask for multi-GPU acceleration?

A. Use dask.distributed.Client() without specifying a scheduler to automatically detect available GPUs. B. Use dask_cuda.LocalCUDACluster() to create a cluster of GPU workers and pass it to the Dask client. C. Assign computation tasks explicitly to CPUs using dask.config.set({'scheduler': 'threads'}) before using GPUs. D. Avoid using dask_cudf and instead rely on standard pandas DataFrames to ensure GPU-accelerated execution.

Correct Answer: B

QUESTION NO: 9
Which of the following Nvidia technologies is commonly used for deploying machine learning models in production environments, enabling scalable deployment and monitoring?

A. Nvidia DeepStream B. Nvidia CUDA C. Nvidia Jetson Nano D. Nvidia Triton Inference Server

Correct Answer: D

NVIDIA-Certified-Professional Accelerated Data Science - NCP-ADS Exam Questions