MassGrid integrates many popular deep learning frameworks such as TensonFlow, PyTorch, and Caffe2 into the MassGrid virtual Docker for users to easily train AI tasks. In MassGrid's huge GPU cluster, in addition to support multiple model training simultaneously , we also integrate optimized decentralized training algorithms such as gradient compression and sparse. With MassGrid's decentralized optimization algorithm, users can easily configure efficient computational training using multiple computing nodes.
