graphcore-research/unit-scaling-demo

Unit Scaling demo and experimentation code

31
/ 100
Emerging

This project offers a way to train language models more efficiently and with better stability, especially when using lower precision computing. It helps machine learning engineers and researchers by taking raw text data, like the WikiText-103 dataset, and applying a technique called Unit Scaling to produce trained character-level language models. The primary users are those working on deep learning model development and optimization.

No commits in the last 6 months.

Use this if you are a machine learning engineer or researcher looking to improve the stability and reduce the computational cost of training deep learning models, particularly character-level language models, by using low-precision techniques.

Not ideal if you are looking for a general-purpose language model for immediate application rather than a specialized tool for research and development into training techniques.

deep-learning-optimization language-model-training low-precision-computing neural-network-research machine-learning-engineering
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 6 / 25
Maturity 16 / 25
Community 9 / 25

How are scores calculated?

Stars

16

Forks

2

Language

Jupyter Notebook

License

MIT

Last pushed

Mar 12, 2024

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/graphcore-research/unit-scaling-demo"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.