Lowaltitudeflight/local-llm-tuning-guide

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.

语言：Python · GitHub 星标：19 · Fork：0

话题：benchmarking · decision-guide · gpu-serving · inference-optimization · kv-cache · llm · llm-deployment · llmops

在 GitHub 上查看原仓库