An end-to-end open source platform for ML inference tasks scheduling in heterogeneous GPU clusters.
For using Dice-server, follow the instruction:
- Clone the repository with option
--recursive
- We use gRPC for communication between server and clients. So, install gRPC from here.
- After installation of gRPC, you may need to modify your bash file. (My shell is zsh so I modified
~/.zshrc
)
echo "export GRPC_INSTALL_DIR=/path/to/grpc/.local" >> ~/.zshrc
echo "export PATH=\"$GRPC_INSTALL_DIR/bin:$PATH\"" >> ~/.zshrc
- Now, you have all of the requirement for your server to be up.
mkdir build; cd build;
cmake ..
make -j all
- Now you only need to run
./server
inbuild
directory. - If you see this output, it means that your server is up.
Server listen on localhost:50051