Home

Awesome

DP-ChemTS

A distributed framework based on Monte Carlo tree search for accelerating molecular discovery. DP-ChemTS is the distributed parallel version of our publish ChemTS library. DP-ChemTS implemented three distributed parallel methods: Distributed leaf parallel, distributed tree parallel and distributed tree parallel with virtual loss(is used to reduce the search overhead).

Requirements

1.mpi4py

2.MPICH

3.Cluster machines

  1. Other requirements for ChemTS.

How to use DP-ChemTS?

  1. Distributed leaf parallel.

cd leaf_parallel_test/simulation1/4core

run: mpiexec -n 4 python mpi_thread_leaf_parallel.py

  1. Distributed tree parallel.

cd tree_parallel_test/simulation1/4core

run: mpiexec -n 4 python mpi_thread_chemts_tree_parallel.py

  1. Distributed tree parallel with virtual loss.

cd virtual_loss_test/simulation1/4core

run: mpiexec -n 4 python mpi_thread_chemts_tree_vl.py