Miatz playground

Fine-Tuning Forge

Slide through cached training runs to find the tune-vs-prompt break-even.

Play it now — free, no account All tools

AIE-110Browser (JS)AI-Coredev · data

The Demystify signature

Fine-tuning is normally a black-box decision made on vendor slides; this lets the learner personally scrub real cached training curves and get a hard, named break-even number instead of a vibe.

How it works

What goes in, what comes out

What it does

Learner explores a library of pre-computed, real LoRA fine-tuning runs (on open-weight base models) by dragging sliders over rank, learning rate, and epoch count, instantly redrawing the cached loss curve and eval-score-over-time for that configuration — no live GPU training, no bill, just real recorded experiment data made explorable. A break-even calculator then compares the all-in fine-tuning cost against simply writing a better prompt for the same task, at a chosen request volume, and names the crossover point.

You bring

A target task/eval set, a chosen base model, and a request-volume estimate for the break-even comparison.

You get

A live-redrawn loss curve and eval-score-over-epoch chart per configuration, and a tune-vs-prompt break-even verdict with the exact request count where fine-tuning starts winning.

You control

Base model, LoRA rank, learning rate, epoch count (all scrubbing pre-computed cached runs), and a request-volume slider for the break-even calculator.

Where it's used

In-module elective extension of AIE-110, positioned as a preview/waitlist teaser for the future L4 fine-tuning track — no live training capability, simulation only.

LoRA/PEFT fine-tuningtune-vs-prompt economicsloss curveseval-score-over-epochtraining cost amortization

Optional · watch & try

Go deeper, elsewhere

Hand-picked public explainers and open tools that complement this one — always optional, never required, never graded.

What is LoRA? Low-Rank Adaptation for finetuning LLMs EXPLAINEDExplains the low-rank matrix trick that makes LoRA cheap LoRA for fine-tuning LLMs explained with exampleWorked example applying LoRA to an actual model LLM Fine Tuning Explained in 8 Minutes: LoRA, QLoRA, DPOFast comparison of fine-tuning method families · ~8 min Hugging Face PEFT — LoRA methodsOfficial reference implementation and guide for LoRA/PEFT fine-tuning LLM Model VRAM Calculator (Hugging Face Space)Estimate GPU memory needed to fine-tune or run a given model at a given quantization

Concepts click when you open the machinery.

Three labs are already live and free — the same hands-on style this playground brings to its module.

Play the free labs Take the DSAT