This tutorial walks through training and deploying a machine learning REST API that predicts cryptocurrency prices in ten steps.
Open source
24 alternativesnaklecha/llama3-from-scratch
This repository walks through an implementation of Llama 3 one matrix multiplication at a time.
About naklecha/llama3-from-scratch
This repository is an educational walkthrough that implements Llama 3 from scratch in a notebook file, one tensor and matrix multiplication at a time. It loads tensors directly from the model file Meta published for Llama 3, so the weights must be downloaded from Meta's official site before the code can run.
The walkthrough covers tokenizing a sample prompt with tiktoken, reading the model configuration, and converting tokens into embeddings while tracking tensor shapes. The author states that the tokenizer is not a BPE implementation and that one embedding step uses a built-in neural network module. The repository is open source.
Key features
- Walks through a Llama 3 implementation one tensor and matrix multiplication at a time
- Loads tensors directly from Meta's Llama 3 model weights file
- Reads model configuration to derive layers, attention heads and vocabulary size
- Tokenizes a sample prompt using tiktoken and Llama 3 special tokens
- Shows token embeddings with tensor shapes, such as 17 tokens by 4096 dimensions
- Includes a notebook file and a requirements file for setup
Best for: Students and developers who want to study how Llama 3 works internally by reading step-by-step code.
Good for
- Explains each step with notes on tensor shapes
- Shows the model configuration values and what they imply about the architecture
- Open source, with a license file included
Watch out for
- Requires separately downloading Meta's model weights before the code can run
- Does not implement a BPE tokenizer; uses tiktoken instead
- Uses a built-in neural network module for the embedding step
naklecha/llama3-from-scratch at a glance
- Category
- Education & Learning › Learn to Code
- Type
- Open-source project
- Pricing
- Not confirmed. Check the site
- Open source
- Yes · Source code
- Language
- English
- Last checked
- Oct 6, 2026 · Online when checked
- Listed
- Oct 5, 2026
naklecha/llama3-from-scratch alternatives
Tools and sites you could use instead, ranked by how closely they match.
A book explains the WebRTC real-time communication framework beyond its APIs.
Open source
24 alternatives
A curated repository helps readers discover machine learning and AI courses published on YouTube.
Open source
24 alternatives
This repository implements JavaScript algorithms and data structures, with explanations and links for further reading.
Open source
24 alternatives
This repository curates learning resources for software engineers, including material for people exploring programming topics.
Open source
24 alternatives
This repository shows how CSS can recreate effects and interfaces without JavaScript.
Open source
24 alternativesPopular in Learn to Code
Well-known listings from the same topic.
Codelabscodelabs.developers.google.comScratchscratch.mit.eduTelegram Bot APIcore.telegram.org/botsStudent Developer Packeducation.github.com/packHacksplaininghacksplaining.comQB64SourceCodeqb64tutorial.com
naklecha/llama3-from-scratch FAQ
Do I need to download anything before running the code?
Yes. The author states that you need to download the Llama 3 weights from Meta's official link before running the file.
Does the repository implement a BPE tokenizer?
No. The author says it does not implement a BPE tokenizer and uses tiktoken instead. The README links to Andrej Karpathy's minbpe project for a BPE implementation.
Is something wrong or out of date?
Tell us and an editor will check it.
Own this website?
Claim this listing
Claim the listing and suggest updates yourself. An editor checks every claim first.