Pinned Loading

1. colibri colibri Public

   Forked from 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. 🐦

   C

2. sonos/tract sonos/tract Public

   Tiny, no-nonsense, self-contained, Tensorflow and ONNX inference

   Rust 3k 273

3. strata strata Public

   C/C++ MoE inference engine for models exceeding local VRAM

   C++ 1