CUDA fork of antirez ds4 that runs giant MoE language models on a single consumer GPU by streaming routed experts from SSD.
Gianni Sanrochman
giannisan
AI & ML interests
Transfromers, LLMs, multimodel, vision, audio
Recent Activity
updated a model about 19 hours ago
giannisan/Hy3-ds4-gguf liked a model 6 days ago
sakamakismile/ThinkingCap-Qwen3.6-27B-NVFP4 updated a collection 6 days ago
NeutronStar