AI0-001 Implementing AI Solutions Practice Question
A company is fine-tuning an LLM for a domain-specific task using LoRA. They have limited GPU memory and need to reduce memory footprint without sacrificing fine-tuning quality. Which approach should they consider?
⚠ Common exam trap
CompTIA often tests the misconception that increasing model capacity (e.g., higher LoRA rank or full fine-tuning) always improves quality, when in fact memory-constrained environments require efficient techniques like QLoRA that balance resource usage and performance.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Use QLoRA with 4-bit quantized base model
QLoRA combines 4-bit NormalFloat quantization of the base model with LoRA adapters, drastically reducing GPU memory usage while preserving fine-tuning quality through techniques like double quantization and paged optimizers. This directly addresses the constraint of limited GPU memory without sacrificing the model's ability to learn domain-specific tasks effectively.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✓
Use QLoRA with 4-bit quantized base model
Why this is correct
QLoRA quantises the frozen base model to 4-bit and trains low-rank adapters, cutting GPU memory substantially while preserving fine-tuning quality. This directly satisfies the stem's limited-memory constraint, unlike standard LoRA, which still loads full-precision base weights.
- ✗
Use a larger batch size to speed up training
Why it's wrong here
A larger batch size multiplies the activations held per step, raising peak memory and often forcing gradient accumulation or smaller micro-batches instead. Larger batches are used to improve throughput and gradient stability when GPU memory already has ample headroom.
- ✗
Fine-tune all layers of the base model
Why it's wrong here
Full fine-tuning updates every weight, so gradients, activations and optimiser states for all parameters must fit in GPU memory — exactly the footprint LoRA avoids by freezing the base model and training small adapter matrices. Full fine-tuning suits maximum-quality domain adaptation when memory is unconstrained.
- ✗
Increase the rank of LoRA adapters
Why it's wrong here
Raising the LoRA rank enlarges the trainable adapter matrices, increasing trainable parameters, gradients and optimiser state, so memory grows rather than shrinks. Higher rank is chosen when the task needs greater adapter capacity and GPU memory is plentiful.
About these practice questions
Courseiva writes every AI0-001 question from scratch — 962 in total, each with an explanation and a wrong-answer breakdown. None are copied from real exams or dumps. Learn why practice questions differ from exam dumps →
JA
Written by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
This AI0-001 practice question is part of Courseiva's free CompTIA certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the AI0-001 exam.