如何进行开发 - Githubissues

# 下载项目
git clone git@github.com:{your_own}/llama.cpp.git

# 进入项目，创建个人开发分支，当前主要开发分支为 ‘npu_support’
cd llama.cpp
git remote add upstream git@github.com:hipudding/llama.cpp.git
git fetch upstream
git checkout -b local_npu_support upstream/npu_support

# 编译
mkdir build 
cd build 
cmake .. -DCMAKE_BUILD_TYPE=debug -DLLAMA_CANN=on && make -j32

# 单算子精度测试
./bin/test-backend-ops test -b CANN0 -o {OP_NAME}
# e.g. 
./bin/test-backend-ops test -b CANN0 -o CONT

# 单算子性能测试
./bin/test-backend-ops perf -b CANN0 -o {OP_NAME}

# 模型推理
./bin/main -m /home/wangshuai/models/hermes_gguf/Hermes-2-Pro-Llama-3-8B-F16.gguf -p "Building a website can be done in 10 simple steps:" -ngl 32 -sm none -mg 0 -t 0
hipudding / llama.cpp

如何进行开发 #4
