CoolFace
Modelpublic

CPU-Hybrid-MoE/DeepSeek-V3-0324-CPU-NUMA4-AMXINT4

sourceHugging Faceupdated 11mo agoView on Hugging Face
0likes17downloads
Model Card

For inference with sglang and kt-kernel: https://lmsys.org/blog/2025-10-22-KTransformers/

This version is packed specifically for NUMA tensor parallel = 4