Fork of FreeToken with DeepSeek-V4.1, vision and speculative decoding (2x3090 numbers inside)

I've been running FreeToken on my 2x3090 box for a while and ended up maintaining a fork of it. Posting it in case it's useful to anyone else here. Quick context if you haven't used it: FreeToken is an edge-native MoE serving engine. It offloads experts to host RAM/NVMe and co-executes on CPU+GPU so

中文全文翻译正在补齐，可先查看下方英文原文。

I've been running FreeToken on my 2x3090 box for a while and ended up maintaining a fork of it. Posting it in case it's useful to anyone else here. Quick context if you haven't used it: FreeToken is an edge-native MoE serving engine. It offloads experts to host RAM/NVMe and co-executes on CPU+GPU so