MANTA GitHub Update 4
Dear Community Members,
Following the previous three updates of MANTA codes to GitHub,
we are uploading the fourth module, i.e. Distributed Network Configuration Module, to GitHub. You can find the relevant codes at https://github.com/MatrixAINetworkMan/MANTA.git
The Distributed Network Configuration Module serves to define the parameters of the distributed MPI framework, which consists of clustered resource management and distributed communication management. The former is for IP mapping of GPU nodes under a cluster. It can deploy distributed computing tasks based on computing chip resource needed. The latter is for the optimisation of configuring the GPU distributed computing. Example of this is the use of IB and NCCL which enables the distributed computing tasks to fully use communication and computation capability of a cluster.
Enjoy the adventure with MATRIX.