#4747th of 65 in Learning and cookbooks

smol-course

Hands-on course on fine-tuning and aligning small language models

Stars
6.8k
License
Apache-2.0
Last commit
Oct 2026
Language
Jupyter Notebook
Stars, last 30 weeks+233
4 Apr 2026: 6,554 stars11 Apr 2026: 6,563 stars18 Apr 2026: 6,571 stars25 Apr 2026: 6,576 stars2 May 2026: 6,581 stars9 May 2026: 6,584 stars16 May 2026: 6,589 stars23 May 2026: 6,592 stars30 May 2026: 6,597 stars6 Jun 2026: 6,607 stars13 Jun 2026: 6,609 stars20 Jun 2026: 6,613 stars27 Jun 2026: 6,620 stars4 Jul 2026: 6,624 stars11 Jul 2026: 6,634 stars18 Jul 2026: 6,641 stars25 Jul 2026: 6,659 stars1 Aug 2026: 6,678 stars8 Aug 2026: 6,694 stars15 Aug 2026: 6,700 stars22 Aug 2026: 6,707 stars29 Aug 2026: 6,714 stars5 Sept 2026: 6,722 stars12 Sept 2026: 6,728 stars19 Sept 2026: 6,738 stars26 Sept 2026: 6,748 stars3 Oct 2026: 6,754 stars10 Oct 2026: 6,764 stars17 Oct 2026: 6,777 stars24 Oct 2026: 6,783 stars31 Oct 2026: 6,787 stars

6,554 on 4 Apr 20266,787 on 31 Oct 2026

From GitHub's daily star counts.

Checked on GitHub.

Overview

A Hugging Face course on aligning small language models, built around SmolLM3 and SmolVLM2. It covers instruction tuning, evaluation, preference alignment with DPO, and vision language models, with reinforcement learning and synthetic data listed as later modules. Exercises are meant to run on most local machines, and learners submit work by pull request and to a leaderboard.

Who it is for: ML engineers learning to fine-tune and align small LLMs and VLMs locally

Strengths

  • Covers SFT, evaluation, DPO and vision language models in one sequence
  • Uses small models (SmolLM3, SmolVLM2) so exercises run on local machines
  • Open to pull requests, with a leaderboard for submitted models
  • Original v1 notebooks are kept in the v1 directory

Weaknesses

  • RL and synthetic data modules were dated Oct and Nov 2025; README does not confirm release
  • README says the course will be re-released on Hugging Face Learn
  • Assumes prior knowledge of PyTorch and the transformers library
  • No exact RAM or GPU figures are given

What it needs

  • GPU optional
  • Needs PyTorch, transformers, Hugging Face Hub
  • Models: SmolLM3, SmolVLM2

Quick answers

Is smol-course open source?

Yes. Its code is on GitHub under the Apache-2.0 license.

Does smol-course need a GPU?

No. It runs on a CPU and can use a GPU when you have one.

Also in Learning and cookbooks

See all 65
Also in Learning and cookbooks
RankProjectScore
1ai-engineering-from-scratchHands-on AI engineering curriculum with 542 lessons across 20 phases76.2k starsMIT89 out of 100
2generative-ai-for-beginners21-lesson course on building generative AI apps in Python and TypeScript122.7k starsMIT81 out of 100
3rag_techniquesCollection of 42 runnable notebooks implementing RAG techniques30.2k starscustom license79 out of 100
#4cs249r_bookOpen textbook and course materials for machine learning systems engineering29.8k starscustom license79 out of 100
#5ai-agents-for-beginners18-lesson course on building AI agents with Microsoft Agent Framework78.6k starsMIT78 out of 100