Supports both imperative and symbolic programmings.
Runs on CPUs or GPUs, on clusters, servers, desktops, or mobile phones
Calculates the gradient automatically for training a model.
Supports distributed training on multiple CPU/GPU machines, including AWS, GCE, Azure, and Yarn clusters.
The well-optimized C++ backend engine parallelize both I/O and computations
MXNet is developed by collaborators from multiple universities and companies. We sincerely thank the following organizations for supporing MXNet and sponsoring its major developers (alphabetical order).