Skip to content

Run frontier MoE models on hardware you already own — pure C, zero deps, experts streamed from disk. Tiny engine, immense model. 🐦

C

Commits on main

fix(omp): size GLM threads from physical cores
bherald committed 3 months ago
948e7b1
374577d
Merge pull request #780 from ThefloorMiner/feat/usage-decay
Vincenzo Fornaro committed 3 months ago
72ddb67
Merge pull request #777 from mgua/main
Vincenzo Fornaro committed 3 months ago
965652f
Merge pull request #779 from JustVugg/feat/fmt8-ue8m0
Vincenzo Fornaro committed 3 months ago
fd5d72e
Merge pull request #778 from JustVugg/feat/st-fp8-dtypes
Vincenzo Fornaro committed 3 months ago
57bf04e
c19206d
test: pin the UE8M0 decode and the dtype table
JustVugg committed 3 months ago
d567972