{"name":"ninfer-fusion-kvmem","repo":"1314521gjy/ninfer-fusion-kvmem","url":"https://midorreal.com/project/1314521gjy-ninfer-fusion-kvmem","repository":"https://github.com/1314521gjy/ninfer-fusion-kvmem","description":"","facts":{"license":{"value":"Apache-2.0","source":"https://github.com/1314521gjy/ninfer-fusion-kvmem","checked":"22 Nov 2026"},"language":{"value":"C++","source":"https://github.com/1314521gjy/ninfer-fusion-kvmem","checked":"22 Nov 2026"},"last_release":{"value":"engine-v0.11.0-kvmem-20261008, 8 Oct 2026","source":"https://github.com/1314521gjy/ninfer-fusion-kvmem/releases","checked":"22 Nov 2026"},"releases":{"value":"1","source":"https://github.com/1314521gjy/ninfer-fusion-kvmem/releases","checked":"20 Nov 2026"},"contributors":{"value":"3","source":"https://github.com/1314521gjy/ninfer-fusion-kvmem/graphs/contributors","checked":"20 Nov 2026"},"archived":{"value":"no","source":"https://github.com/1314521gjy/ninfer-fusion-kvmem","checked":"22 Nov 2026"}},"stars":603.0,"writeup":{"what_it_is":"A Windows and NVIDIA GPU inference project for long-context language models, built around a KV-cache ring that keeps a device-side pool smaller than the logical context. Pages not kept on the device are stored in host memory and retrieved when needed. It is a downstream derivative of NInfer and ships source, reproduction notes and verification tools.","audience":"Developers testing long-context LLM inference on Windows NVIDIA GPUs","claims":[{"kind":"specific","claim":"Device KV pool can be smaller than the logical context length","excerpt":"让设备 KV 池可以小于逻辑上下文","status":"not_checked"},{"kind":"specific","claim":"Supports sm_86, sm_89 and sm_120a builds of the engine","excerpt":"sm86 / sm89 / sm120a 三份引擎由同一份源码各自编译","status":"not_checked"},{"kind":"specific","claim":"Speculative decoding with MTP reports an acceptance rate of 77-92% on the author's machine","excerpt":"本机实测 `mtp accepted 77–92%`","status":"not_checked"},{"kind":"broad","claim":"Inference is fast and easy to use","excerpt":"面向 Windows / NVIDIA GPU 的长上下文推理实验","status":"too_broad"}],"alternatives":[],"written_by":"AI, from the project's README","date":"2026-10-25"},"why_now":null,"owner_supplied":null,"score":{"verdict":"overhyped","hype":32,"reality":2,"gap":30,"momentum":0.3,"confidence":60,"version":"score-2.0","rule":"verdict-2.0","calculated_at":"2026-11-22T07:00:00+00:00","verdict_source":"rule"},"early_read":null,"changes":[{"date":"2026-10-08T08:56:09+00:00","text":"Latest release engine-v0.11.0-kvmem-20261008","source":"https://github.com/1314521gjy/ninfer-fusion-kvmem/releases"},{"date":"2026-10-04T12:04:52+00:00","text":"Repository created","source":"https://github.com/1314521gjy/ninfer-fusion-kvmem"}],"cite":"ninfer-fusion-kvmem is a GitHub project at https://github.com/1314521gjy/ninfer-fusion-kvmem written mainly in C++.","attribution":"Data from Mid or Real, https://midorreal.com/project/1314521gjy-ninfer-fusion-kvmem, checked 22 November 2026. Attribution with a link is required.","still_real":null}