Portable, scalable and reliable distributed machine learning.

Wormhole is a place where DMLC projects works together to provide scalable and reliable machine learning toolkits that can run on various platforms

• Portable:
  • Supported platforms: YARN, MPI and Sungrid Engine
  • Planned: docker support
• Rich support of Data Source
  • All projects can read data from HDFS, S3 or local filesystem
• Scalable and Reliable

• Boosted Trees (GBDT): XGBoost: eXtreme Gradient Boosting
• L-BFGS based linear solver
• Asynchrouns SGD linear solver

• copy make/config.mk to root folder
• modify according to your settings
• type make or make name-of-tool-you-want

• make sure dmlc-core exist in root folder
  • type make dmlc-core to get it
• Use the submission script in dmlc-core/tracker to submit job to the platform of your choice

• We believe that we can create machine learning tools that are portable and works with each other.
• Contributing of machine learning projects, tutorials and to core dmlc projects are welcomed.
  • All machine learning projects can depends on dmlc-core, rabit or parameter-server

• learn contains simple but powerful learning tools in wormhole
• repo is used to clone other DMLC repos that wormhole can depend on
• Depending DMLC Libraries
  • dmlc-core gives the core modules of most DMLC projects.
  • rabit provides reliable BSP Allreduce communication.
  • parameter-server provides asynchronize parameter server abstraction.