google/paxml
Pax is a Google-developed JAX-based machine learning framework for training large-scale models with advanced parallelism and configurability.
Not currently ranked — collecting fresh signals.
star history
Pax is a framework built on JAX for configuring and running machine learning experiments at scale. It provides advanced parallelization capabilities and has achieved industry-leading model flop utilization rates. The framework is particularly oriented toward training large language models as evidenced by topics like GPT, LLM, and large-language-models, and supports distributed training across Google Cloud TPU infrastructure.
Frequently asked
- What is google/paxml?
- Pax is a Google-developed JAX-based machine learning framework for training large-scale models with advanced parallelism and configurability.
- Is paxml open source?
- Yes — google/paxml is open source, released under the Apache-2.0 license.
- What language is paxml written in?
- google/paxml is primarily written in Python.
- How popular is paxml?
- google/paxml has 555 stars on GitHub.
- Where can I find paxml?
- google/paxml is on GitHub at https://github.com/google/paxml.