Ninfer-V100-16G 实现单卡 68t/s

Name: Ninfer-V100-16G 实现单卡 68t/s
Uploaded: 2026-10-04T14:37:30.000Z
Duration: 3 min 33 s
Channel: 崩坏的领航员
Description: 最近尝试在v100-16g上使用ninfer，fork了geoffwatts/ninfer-v100: High-performance single-GPU inference for selected model checkpoints and GPUs.之后让ai改了几轮，适配了q2的模型，比llama.cpp快不少 后续会更新到 CollapseNav/ninfer-v100-16g: 针对V100-16g调整过的ninfer

2698

2026-10-04 22:37:30

个人观点，仅供参考

最近尝试在v100-16g上使用ninfer，fork了geoffwatts/ninfer-v100: High-performance single-GPU inference for selected model checkpoints and GPUs.之后让ai改了几轮，适配了q2的模型，比llama.cpp快不少后续会更新到 CollapseNav/ninfer-v100-16g: 针对V100-16g调整过的ninfer

最近的ai可太好玩了

顶部