google/paxml

Pax is a Jax-based machine learning framework for training large scale models. Pax allows for advanced and fully configurable experimentation and parallelization, and has demonstrated industry leading model flop utilization rates.

63
/ 100
Established

Built on Praxis (a configuration-based layer abstraction), Pax uses JAX's `pjit` (SPMD) and `pmap` primitives to express complex distributed training strategies across TPU Pods and GPU clusters without explicit communication code. The framework emphasizes declarative experiment configuration, enabling researchers to specify model architecture, data pipeline, and parallelization strategy through composable Python configs that can be version-controlled and reproduced.

550 stars. Actively maintained with 5 commits in the last 30 days. Available on PyPI.

Maintenance 16 / 25
Adoption 10 / 25
Maturity 18 / 25
Community 19 / 25

How are scores calculated?

Stars

550

Forks

70

Language

Python

License

Apache-2.0

Category

llm-fine-tuning

Last pushed

Mar 12, 2026

Commits (30d)

5

Dependencies

20

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/llm-tools/google/paxml"

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