CoursesFree Quizzes
NVIDIAFine-Tuning

Parameter Efficient & Lora — NCP-GENL Practice Question

A representative NVIDIA-Certified Professional Generative AI LLMs (NCP-GENL) exam question on Parameter Efficient & Lora. Work through it below, then read why each option is right or wrong.

This question is part of the free NVIDIA-Certified Professional Generative AI LLMs practice test.

Short answer

The correct answer is A. Use parameter-efficient fine-tuning (LoRA or adapters) and transfer only the small adapter artifact (~50 MB) across the regional security boundary, keeping the base model local to each region to minimize cross-region data transfer..

Parameter-efficient fine-tuning (LoRA or adapters) is the optimal choice for security-boundary constraints because it minimizes cross-region data transfer. By transferring only the small adapter artifact (~50 MB) instead of the full model (~13 GB), the approach reduces bandwidth consumption, exposure time, and security risk across the regional firewall. The base model can be deployed locally to each region, and only the lightweight adapter is transferred, maintaining consistency while respecting network security boundaries. Full fine-tuning requires transferring the entire model across the boundary, increasing security exposure. Continuous pretraining sacrifices consistency. Quantization reduces accuracy unnecessarily when parameter-efficient methods already solve the problem.

The Question

A manufacturing plant is deploying an LLM to generate safety recommendations and equipment-maintenance schedules. The plant operates on a private, air-gapped network with no internet access. A regional disaster has damaged the primary data center, and the plant must restore the LLM to a secondary site in a different region. The secondary site has limited network connectivity between regions due to firewall and security-boundary restrictions. The team must choose a fine-tuning method that minimizes the data that must be transferred between regions during recovery. Which security boundary consideration is most critical?

AUse parameter-efficient fine-tuning (LoRA or adapters) and transfer only the small adapter artifact (~50 MB) across the regional security boundary, keeping the base model local to each region to minimize cross-region data transfer.Correct
BUse domain adaptation with continuous pretraining on local data in the secondary region, avoiding cross-region transfer but sacrificing consistency with the primary-region model.
CUse full fine-tuning and transfer the entire fine-tuned model (~13 GB for a 7B model) across the regional network, accepting high bandwidth consumption and potential security exposure.
DDeploy a lightweight quantized model to reduce artifact size, then transfer it across the regional boundary, accepting potential accuracy loss from quantization.

Why A is correct

Parameter-efficient fine-tuning (LoRA or adapters) is the optimal choice for security-boundary constraints because it minimizes cross-region data transfer. By transferring only the small adapter artifact (~50 MB) instead of the full model (~13 GB), the approach reduces bandwidth consumption, exposure time, and security risk across the regional firewall. The base model can be deployed locally to each region, and only the lightweight adapter is transferred, maintaining consistency while respecting network security boundaries. Full fine-tuning requires transferring the entire model across the boundary, increasing security exposure. Continuous pretraining sacrifices consistency. Quantization reduces accuracy unnecessarily when parameter-efficient methods already solve the problem.

Why the other options are wrong

Option B: Use domain adaptation with continuous pretraining on local data in the secondary region, avoiding cross-region transfer but sacrificing consistency with the primary-region model.

Option B does not satisfy the requirement in the scenario. Review the explanation above: the correct choice (A) is the only one that fully meets every constraint stated in the question.

Option C: Use full fine-tuning and transfer the entire fine-tuned model (~13 GB for a 7B model) across the regional network, accepting high bandwidth consumption and potential security exposure.

Option C does not satisfy the requirement in the scenario. Review the explanation above: the correct choice (A) is the only one that fully meets every constraint stated in the question.

Option D: Deploy a lightweight quantized model to reduce artifact size, then transfer it across the regional boundary, accepting potential accuracy loss from quantization.

Option D does not satisfy the requirement in the scenario. Review the explanation above: the correct choice (A) is the only one that fully meets every constraint stated in the question.

Key idea: Parameter Efficient & Lora

Parameter-efficient fine-tuning (LoRA or adapters) is the optimal choice for security-boundary constraints because it minimizes cross-region data transfer. By transferring only the small adapter artifact (~50 MB) instead of the full model (~13 GB), the approach reduces bandwidth consumption, exposure time, and security risk across the regional firewall. The base model can be deployed locally to each region, and only the lightweight adapter is transferred, maintaining consistency while respecting network security boundaries. Full fine-tuning requires transferring the entire model across the boundary, increasing security exposure. Continuous pretraining sacrifices consistency. Quantization reduces accuracy unnecessarily when parameter-efficient methods already solve the problem. On the NCP-GENL exam, questions in the "Fine-Tuning" domain test whether you can map a scenario's constraints to the right choice. Read the requirement carefully, eliminate options that violate any single constraint, and pick the one that satisfies all of them with the least operational overhead.

Ready to see how you'd score?

Take the free practice quiz and find out which NVIDIA-Certified Professional Generative AI LLMs domains you need to focus on. No signup required.

Take the free quizSee all 222+ practice questions

Practice 5 similar questions

Same cert, same or adjacent domain. Use these after reviewing the explanation.

1Multi Head Attention & Attention Specialization — NCP-GENL Practice QuestionLLM Architecture2Nemo Guardrails & Jailbreak Detection — NCP-GENL Practice QuestionSafety, Ethics, and Compliance3Chain-of-Thought & Output Control — NCP-GENL Practice QuestionPrompt Engineering4Nemo Curator & Semantic Deduplication — NCP-GENL Practice QuestionData Preparation5Quantization & Memory Optimization — NCP-GENL Practice QuestionModel Optimization

Related NVIDIA-Certified Professional Generative AI LLMs Practice Questions

Multi Head Attention & Attention Specialization — NCP-GENL Practice Question→Nemo Guardrails & Jailbreak Detection — NCP-GENL Practice Question→Chain-of-Thought & Output Control — NCP-GENL Practice Question→

Official sources used

Verify program and exam details at the source before scheduling.

docs.nvidia.comLLM: Index.htmldocs.nvidia.comOptimized: Index.html