This Confluence has been LDAP enabled, if you are an ASF Committer, please use your LDAP Credentials to login. Any problems file an INFRA jira ticket please.

Page tree
Skip to end of metadata
Go to start of metadata

Last updated: initial draft

Last discussion thread(s):

The MXNet roadmap is a list of high level ideas and desired timelines. The roadmap is not a commitment on schedules and subject to change. Typically the roadmap will be updated quarterly based on new insights, learnings and user feedback.

In the future the project might consider to adopt a process for consideration of new suggestions similar to Apache Spark Improvement Proposals.

Projects to be developed as part of Apache MXNet ecosystem can be included.

Q3 2018: Focus on Quality and Technical Debt

  • Resolution of github issues: bugs, tests, builds, installation
  • Translate Documentation
  • Documentation of inference thread support and limitations
  • High quality support for MKL (incl. MKL-DNN)
  • Subgraph API for integrating backend accelerators with MXNet
  • quantization flow for INT8 with MKL-DNN (suggested by Patrick patric.zhao@intel.com)
  • Integration with Tensor RT (experimental in Q2)
  • Gluon CV, NLP toolkit
  • Increase example and tutorial coverage for various application domains - “MXNet by Example”
  • Performance profiling and improvements
  • Language API improvements - what is the community interest?: R (training), Java (wrapped around Scala?, inference), C++ (inference)

Q4 2018: Focus on feature gaps

  • Resolution of github issues: performance, operators, feature requests
  • Resolution of github issues: R, Gluon, C++, Python
  • Scalability improvement for distributed processing
  • Distributed training for Gluon CV, NLP toolkit
  • Community contributions: array creation routines
  • CI & validation for MXNet examples 
  • Increase example and tutorial coverage for various application domains - “MXNet by Example”
  • Android SDK for mobile devices

Q1 2019: Focus on Language AP

  • Increase example and tutorial coverage for various application domains - “MXNet by Example”
  • Support for low-bit precision inferencing
  • IoT device support for inferencing
  • MKL-DNN RNN API supports (patric.zhao@intel.com)
  • Support for MacOS High Sierra and Mojave

Q2 2019: TBD

  • Increase example and tutorial coverage for various application domains - “MXNet by Example”

Future: Longterm ideas

  • Model interpretability
  • No labels