# angelos-p/llm-from-scratch

Repository: https://github.com/angelos-p/llm-from-scratch
Canonical: https://ross.abutalabs.com/products/llm-from-scratch
License Family: other
Last push: 2026-05-05T09:22:14+00:00

## Health v2 (maintenance only)
Score: 50/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 80, release rhythm 35, longevity 10
- inputs: {"age_days": 151, "days_push": 120, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, young, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3380, forks 367 (observed 2026-08-28T04:07:58.612978+00:00)

## What it is
A hands-on workshop repository where learners write every component of a GPT training pipeline (tokenizer, transformer, training loop, generation) from scratch in PyTorch. It targets a ~10M parameter model that trains on a laptop in under an hour, inspired by Karpathy's nanoGPT.

## Use cases
- learn how transformers and GPT work by building one from scratch
- train a small language model on my laptop
- understand tokenization, attention, and training loops hands-on
- teach an LLM workshop session
- generate Shakespeare-like text with a model I wrote myself
- get started with LLMs without prior ML experience

## When to choose
- you want to deeply understand GPT internals by writing every piece yourself
- you have no GPU cluster and want a model that trains on a laptop in under an hour
- you're a beginner with Python skills but no ML background
- you want a single-session guided workshop format

## When to avoid
- you need to train production-scale or GPT-2-sized models
- you want a ready-made library or framework rather than guided exercises
- you need fine-tuning, RLHF, or instruction-tuning pipelines
- you want prebuilt model weights or inference tooling

## Facets
- artifact type: learning-resource
- maturity: active
- function: llm-training, machine-learning, deep-learning
- domain: large-language-models, deep-learning, machine-learning, education, tutorials
- platform: python, windows, cross-platform
- tags: gpt, transformer, workshop, nanogpt, pytorch, hands-on, from-scratch, education, macos, linux

## Member repositories
- angelos-p/llm-from-scratch (main) score 50

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:58.612978+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 2026-08-29T18:39:41.518895+00:00, confidence not recorded.
  - readme: https://github.com/angelos-p/llm-from-scratch (fetched 2026-08-28T04:07:58.612978+00:00, sha 8e1d64f4675f)
- Data as of 2026-08-30T08:39:29.467469+00:00.