# huggingface/smol-course

A course on aligning smol models.

Repository: https://github.com/huggingface/smol-course
Canonical: https://ross.abutalabs.com/products/smol-course
Language: Jupyter Notebook
License: Apache-2.0
License Family: permissive
Last push: 2026-08-20T14:00:51+00:00

## Health v2 (maintenance only)
Score: 66/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 98, release rhythm 35, longevity 46
- inputs: {"age_days": 646, "days_push": 13, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 6731, forks 2276 (observed 2026-08-28T04:09:48.131938+00:00)

## What it is
A free, open-source course from Hugging Face on aligning small language models (SmolLM3, SmolVLM2) to specific use cases, covering instruction tuning, evaluation, preference alignment like DPO, and vision-language models. It is designed to run on modest local hardware with minimal GPU requirements and includes exercises, a leaderboard, and community peer review.

## Use cases
- learn how to fine-tune a small language model
- align an LLM to a custom domain with DPO
- run LLM fine-tuning on a local machine without paid GPUs
- evaluate fine-tuned language models with benchmarks
- adapt multimodal vision-language models
- generate synthetic training data for a custom domain

## When to choose
- you want a hands-on, practical introduction to LLM alignment and fine-tuning
- you have limited compute and need everything to run locally
- you want to work with SmolLM3 or SmolVLM2 specifically
- you prefer community-reviewed, open course material

## When to avoid
- you need production-scale training of large models on multi-GPU clusters
- you want a reference tool or library rather than educational material
- you need coverage of topics still marked unreleased, like reinforcement learning or synthetic data

## Facets
- artifact type: learning-resource
- maturity: active
- function: machine-learning, llm-training, rag, prompt-engineering
- domain: large-language-models, machine-learning, education, tutorials
- platform: python, cross-platform
- tags: fine-tuning, alignment, dpo, instruction-tuning, small-language-models, vision-language-models, hugging-face, jupyter-notebooks, course

## Member repositories
- huggingface/smol-course (main) score 66

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:48.131938+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:42:48.598850+00:00, confidence not recorded.
  - readme: https://github.com/huggingface/smol-course (fetched 2026-08-28T04:09:48.131938+00:00, sha 2cdbd16781ce)
- Data as of 2026-08-30T08:39:29.467469+00:00.