Lightlines
By wshobson

finetuning-method-selection

wshobson2.1k installs

Choose whether to train, and which method fits.

What it does

Decides whether training is appropriate and recommends a fine-tuning method and base-model size class.

When to use it

Use before starting fine-tuning, when comparing training with RAG or prompting, or when choosing and sizing a training method.

How to use it

Give it a proposed training problem and the available data shape; it recommends whether to train and which method fits.

Uses

Good to know

  • cannot be undone: references/model-catalog.md
  • edits its own instructions

Latest review

What you need · 5

An evaluation harness must exist before a training method is selected.

Hands off SFT execution to this skill after selecting the method; installation is not stated as necessary for the routing decision itself.


Access · 3

This skill

references/model-catalog.md

Write

Updates its model catalog (irreversible)

Unsloth supported-models page

Read

Checks Unsloth model support