NCA-GENL Question 169
Select 3A company plans to deploy a generative AI large language model (LLM) to provide customer support via a chatbot. The LLM requires high computational power for inference and low-latency access to training data for periodic fine-tuning. Which combination of system components best meets these requirements?
- A
High-performance GPUs such as NVIDIA A100 or H100 for accelerated inference and training
- B
High-capacity HDD storage for storing the training dataset
- C
High-speed NVMe SSDs for low-latency access to the training dataset
- D
A CPU-based server cluster optimized for general-purpose computation
- E
A cloud platform that supports scaling of GPU resources on demand
Show answer and explanation
Correct answers: A, C, E
Explanation
Deploying a generative AI LLM for customer support requires a combination of system components that address both computational and data access needs. High-performance GPUs like NVIDIA A100 or H100 are essential for efficient inference and training. High-speed NVMe SSDs ensure low-latency access to training data, crucial for fine-tuning. Additionally, leveraging a cloud platform with on-demand GPU scaling provides the flexibility to meet varying workload demands, making these components ideal for this scenario.
- A. Correct.
High-performance GPUs such as NVIDIA A100 or H100 are critical for handling the computational demands of LLM inference and fine-tuning efficiently. These GPUs are optimized for deep learning workloads.
- B. Incorrect.
High-capacity HDD storage might provide the necessary storage space, but it lacks the low-latency access required for efficient training and fine-tuning of LLMs.
- C. Correct.
High-speed NVMe SSDs provide the required low-latency access to the training dataset, ensuring faster data throughput during fine-tuning processes.
- D. Incorrect.
A CPU-based server cluster is not sufficient for the intensive computational requirements of LLM inference and fine-tuning, as these tasks are better suited for GPUs.
- E. Correct.
A cloud platform that supports scaling of GPU resources on demand ensures flexibility and the ability to meet varying computational needs efficiently, especially during high-demand periods.