tardellirs/gemma3moe-builder
Sparse-upcycle a dense Gemma-3-270m into a Mixture-of-Experts (gemma3moe) and serve it on llama.cpp (CPU). Upcycle, train, inspect routing, GGUF.
⭐ 43
⑂ 0
Python
NOASSERTION
· 26 天前推送
43
Watchers
0
贡献者
0
Commits
0
Releases
0
Open Issues
26 天前
最近推送
原文
中文
暂无 README