Lowaltitudeflight/local-llm-tuning-guide
在 GitHub 打开 ↗A decision library for running local LLMs on multi-GPU machines without NVLink: parallelism, communication, quantization, speculative decoding, KV cache. Every conclusion carries its evidence grade and reproducible conditions.
Star29
24h 新增—
7 天新增—
Fork0
Issue0
语言Python
项目简介
A decision library for running local LLMs on multi-GPU machines without NVLink: parallelism, communication, quantization, speculative decoding, KV cache. Every conclusion carries its evidence grade and reproducible conditions.
Star 趋势
话题标签
benchmarkingdecision-guidegpu-servinginference-optimizationkv-cachellmllm-deploymentllmopslocal-llmmodel-parallelismmulti-gpunvidianvlinkpcieperformance-tuningquantizationself-hostedspeculative-decodingtensor-parallelismvllm
仓库信息
| License | MIT |
| 创建时间 | 2026-10-16 |
| 最后推送 | 1 天前 |
| 热度分 | 0.0 |