WARNING: THIS IS NOT AN APACHE SOFTWARE FOUNDATION RELEASE OF MXNET AS IT PREDATES MXNET JOINING THE APACHE SOFTWARE FOUNDATION

• Move symbolic API to NNVM @tqchen
  • Most front-end C API are backward compatible
  • Removed symbolic api in MXNet and relies on NNVM
• New features:
  • MXNet profiler for profiling operator level executions
  • mxnet.image package for fast image loading and processing
• Change of JSON format
  • param and attr field are merged to attr
  • New code is backward compatible can load old json format
• OpProperty registration now is deprecated
  • New operators are encouraged to register their property to NNVM op registry attribute
• Known features removed limitations to be fixed
  • Bulk segment execution not yet added.
• Miscellaneous
  • sgd and adam optimizer are now implemented with a single imperative call. They should be as fast and memory efficient as cc optimizers. ccsgd is now deprecated and redirects to sgd.
  • Layout support is added. Use mx.io.DataDesc(..., layout='NHWC') in provide_data to specify data layout. use mx.sym.YourSymbol(..., __layout__='NHWC') to specify output layout. layout option is now available for Convolution layer.
  • element_mask is removed. Please use src*mask.reshape((mask.size, 1, 1, ..., 1)) directly as binary ops now support broadcasting.
  • sum_axis, max_axis, and min_axis are deprecated. Please use mx.nd.max(src, axis=n) instead.
  • symbol attributes are now limited to ctx_group, lr_mult, wd_mult, force_mirroring. All other custom attributes need to be in xxx format (start and end with double underscore) or error will be triggered during attribute parsing.