I have written an article about how to build a distributed training cluster on Raspberry Pie 4 using the distributed training system that comes with TF2.0. However, there is a drawback: we need to start the training program at each node, and the distributed training will only work after all the nodes are started. MPI is mainly used in the field of supercomputing. Building MPI cluster on Raspberry, firstly, it can be used to learn distributed computing on supercomputing, and secondly, it can…

Tensorflow Converter can convert models to FlatBuffers format. FlatBuffers is a cross-platform serialization tool where structured data is stored in binary form and can be used in the following ways The performance on the microcontroller is memory saving.

Continuing from the previous article on NFS file systems, the current requirement is to be able to control the applications running in the cluster. For example, if we need to run parallel cross-node program A, we need to schedule and stop the running program B. This requires building a messaging system that can perform operational tasks based on message drivers.

目前正学习如何在树莓派4集群上结合gromacs+mpich的分子动力学模拟，所以需要搭建一款分布式存储系统，又由于树莓派性能的限制，搭建OpenEBS会比较浪费计算资源，最后就采用轻量级的NFS来完成文件系统的共享和存储。

The Rosetta@home project computationally generated and used in animal experiments an immune protein for SARS-COV-2 that could serve to protect animals from a lethal new coronavirus, which is still in the experimental phase and is being optimized.

目前自己采购了35块树莓派4core4G用来计算Rosetta@home，等到COVID-2019过去之后，会用这些开发板测试TF项目，因为TF的生态链齐全，所以在工业上会考虑使用TF来做为最终的产品技术使用方案。