DAY 71 / 210

Intro to Parameter-Efficient Fine-Tuning

Phase 2 begins by establishing why full fine-tuning is rarely viable for startup-scale models; learners must internalize the efficiency trade-offs that will govern every later experiment. This foundation directly informs the rate-limiter and brief-form patterns already present in the Maku codebase.

⏱ 45 min target📝 2 quiz Qs

Resources

readingarXiv
LoRA: Low-Rank Adaptation of Large Language Models
Abstract + Sections 1-2
20 min
readingHugging Face
PEFT: Parameter-Efficient Fine-Tuning
Overview + Quickstart
15 min

Deliverable

Journal entry containing a one-paragraph fine-tuning plan that references the existing rate-limiter logic

Quiz · 2 questions

1. Why does LoRA freeze the base weights instead of updating them?

To reduce VRAM usageTo increase model sizeTo avoid all regularizationTo enable full retraining

2. Name one concrete risk of applying full fine-tuning to the brief-generation endpoint.

Journal

Time spent (minutes)

Blockers

Commit / PR links (one per line)