Observed arrival · 2026-10-13
mojolearn: bitwise-reproducible GPU machine learning
An open-source Mojo project for reproducing machine-learning training and inference results bit-for-bit across supported GPU hardware.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Machine-learning researchers and GPU software developers
- Worth noticing
- The documented experiment compares six hardware segments and reports matching state through 5,000 optimizer steps.
Field notes
The reported experiment uses a 162,147,840-parameter model with GPT-3 Small’s depth and width, modern Llama-style blocks, and FP32 throughout. Each route processes 2.62 billion FineWeb-Edu tokens, switching hardware across six segments; the page says the model card includes architecture, quality evaluations, limitations, weights, and route evidence. It also links a support matrix, release history, and verification guide.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue