HN · GitHub Signals

OPEN SOURCE, IN FOCUS

skewadam

nuemaan/skewadam

Tiered optimizer state allocation for memory-efficient MoE training. Cuts optimizer memory by 97.4%, outperforming AdamW/Muon/Lion while fitting a 6.78B MoE on a single 40GB GPU.

Explore on GitHub ↗

Saved projects stay in this browser. No account needed.

TAKE A CLOSER LOOK

Start with the source.

The repository is the place to check installation steps, supported platforms, the license, and current maintenance before choosing a tool.

Read the documentation on GitHub ↗

Keep exploring

Find projects in the same areas, or search the archive for a different approach.

ANOTHER GOOD FIND

Explore related projects

View the archive ↗

Connected by topics and descriptions.

the-school-of-ai

llm

LightningLM 0.1V — Reference training pipeline for the LightningLM family. 2B dense seed → 5B MoE → 9B MoE → 120B sparse MoE through TurboQuant-PreTraining on a single eight-GPU node. Companion code for *Reversible Foundations*.

Take a look →
dicklesworthstone

franken_ocr

Pure-Rust, CPU-only OCR engine for Baidu Unlimited-OCR (a DeepSeek-OCR-derived 3B MoE VLM). One fixed model, custom int4/int8 kernels, no ML framework, no GPU. Pre-Phase-0 scaffold.

Take a look →
justvugg

colibri

Run GLM-5.2 (744B MoE) on a 25GB-RAM consumer machine — pure C, zero deps, experts streamed from disk. Tiny engine, immense model. 🐦

Take a look →