graphcore-research/jax-scalify

JAX Scalify: end-to-end scaled arithmetics

31
/ 100
Emerging

This tool helps machine learning engineers and researchers accelerate the training and inference of deep neural networks. It takes your existing JAX-based neural network models and training data, then outputs a modified model that can run efficiently using lower precision numbers like BF16, FP16, or even FP8. This allows you to develop and deploy models faster and with less computational resources.

No commits in the last 6 months. Available on PyPI.

Use this if you are a deep learning practitioner working with JAX and want to speed up model training and inference by leveraging low-precision arithmetic without complex manual scaling implementations.

Not ideal if you are not using JAX, or if your primary goal is not related to optimizing neural network performance through precision scaling.

deep-learning neural-network-training model-optimization machine-learning-engineering numerical-precision
Stale 6m
Maintenance 0 / 25
Adoption 6 / 25
Maturity 25 / 25
Community 0 / 25

How are scores calculated?

Stars

18

Forks

Language

Python

License

Apache-2.0

Category

llm-fine-tuning

Last pushed

Oct 30, 2024

Commits (30d)

0

Dependencies

5

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/transformers/graphcore-research/jax-scalify"

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