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.
Paxml helps machine learning engineers efficiently train very large-scale deep learning models, particularly large language models like GPT-3, on Google Cloud TPUs. It takes your model configurations and training datasets, and outputs trained models along with performance metrics like loss curves and perplexity. This framework is for machine learning practitioners and researchers working with massive datasets and models requiring significant computational resources.
550 stars. Actively maintained with 7 commits in the last 30 days. Available on PyPI.
Use this if you are a machine learning engineer or researcher training large-scale deep learning models, especially large language models, and need to optimize their performance and computational efficiency on Google Cloud TPUs.
Not ideal if you are working with smaller models, do not have access to Google Cloud TPUs, or prefer a framework that offers broader GPU support out-of-the-box (though an NVIDIA-optimized version exists for GPUs).
Stars
550
Forks
70
Language
Python
License
Apache-2.0
Category
Last pushed
Mar 12, 2026
Commits (30d)
7
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.
Related tools
axolotl-ai-cloud/axolotl
Go ahead and axolotl questions
JosefAlbers/PVM
Phi-3.5 for Mac: Locally-run Vision and Language Models for Apple Silicon
iamarunbrahma/finetuned-qlora-falcon7b-medical
Finetuning of Falcon-7B LLM using QLoRA on Mental Health Conversational Dataset
h2oai/h2o-wizardlm
Open-Source Implementation of WizardLM to turn documents into Q:A pairs for LLM fine-tuning
WangRongsheng/Aurora
The official codes for "Aurora: Activating chinese chat capability for Mixtral-8x7B sparse...