what can MassGrid do？
MassGrid integrates the most popular deep learning frameworks such as TensonFlow, PyTorch, and Caffe2 into the MassGrid virtual Docker so users can easily train AI tasks. MassGrid’s mass GPU cluster not only integrates optimized decentralized training algorithms such as gradient compression and sparse, but also supports simultaneous training of multiple models . With MassGrid’s decentralized optimization algorithm, users can easily configure efficient computational training using multiple computing nodes.
MassGrid将TensonFlow，PyTorch，Caffe2等多种流行的深度学习框架集成到MassGrid虚拟Docker中，供用户方便地进行AI任务的训练。在MassGrid庞大的GPU集群，除了支持用户多个模型同时训练外，我们也集成了优化的分布式训练算法，如梯度压缩和稀疏化等。利用MassGrid的分布式优化算法，用户可以简单地配置使用多个计算节点进行高效地模型训练。