Skip to content

Repository files navigation

lhm

Local high model.
Focus on user local device to run one model.

Inspired by llama.cpp

  1. Fork from llama.cpp
  2. Trim some old model
  3. Use third party library
  4. Just support one model to run
  5. Merge some code from llama.cpp by time

build

linux cpu

cmake -B build -DGGML_CUDA=OFF
cmake --build build --config Release -j 8

linux cuda

cmake -B build -DGGML_CUDA=ON
cmake --build build --config Release -j 8

mac

cmake -B build -DGGML_CUDA=OFF -DGGML_METAL=ON
cmake --build build --config Release -j 8

windows

build with mooncake

mooncake as kvcache

run

./bin/llm_cli --model /modelpath/model.gguf --log_level info

About

Local high model, run in local device.

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages