jaepil/geometric-adam
A Ray Tracing-Inspired Approach to Neural Network Optimization
This optimization algorithm helps machine learning engineers train large neural network models more reliably. It takes your model's parameters and trains them with an innovative 'ray tracing' approach, preventing common training failures and producing a more accurate, stable model. It's designed for machine learning researchers and engineers who work with deep learning models, especially large transformers, who struggle with models diverging during long training runs.
No commits in the last 6 months.
Use this if your large neural network models frequently diverge or fail during training, and you need a more stable and robust way to achieve high performance over many training epochs.
Not ideal if you are looking for a fully production-ready, off-the-shelf solution, as this project is an active research exploration rather than a finished product.
Stars
17
Forks
—
Language
Python
License
MIT
Category
Last pushed
Jun 11, 2025
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/transformers/jaepil/geometric-adam"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
ModelCloud/GPTQModel
LLM model quantization (compression) toolkit with hw acceleration support for Nvidia CUDA, AMD...
intel/auto-round
🎯An accuracy-first, highly efficient quantization toolkit for LLMs, designed to minimize quality...
pytorch/ao
PyTorch native quantization and sparsity for training and inference
bodaay/HuggingFaceModelDownloader
Simple go utility to download HuggingFace Models and Datasets
NVIDIA/kvpress
LLM KV cache compression made easy