Release Notes - SINGA - Version singa-incubating-1.1.0

SINGA is a general distributed deep learning platform for training big deep learning models over large datasets.

This release includes following features:

  * Core components
      * [SINGA-296] Add sign and to_host function for pysinga tensor module

  * Model components
      * [SINGA-254] Implement Adam for V1
      * [SINGA-264] Extend the FeedForwardNet to accept multiple inputs
      * [SINGA-267] Add spatial mode in batch normalization layer
      * [SINGA-271] Add Concat and Slice layers
      * [SINGA-275] Add Cross Entropy Loss for multiple labels
      * [SINGA-278] Convert trained caffe parameters to singa
      * [SINGA-287] Add memory size check for cudnn convolution

  * Utility functions and CI
      * [SINGA-242] Compile all source files into a single library.
      * [SINGA-244] Separating swig interface and python binding files
      * [SINGA-246] Imgtool for image augmentation
      * [SINGA-247] Add windows support for singa
      * [SINGA-251] Implement image loader for pysinga
      * [SINGA-252] Use the snapshot methods to dump and load models for pysinga
      * [SINGA-255] Compile mandatory depedent libaries together with SINGA code
      * [SINGA-259] Add maven pom file for building java classes
      * [SINGA-261] Add version ID into the checkpoint files
      * [SINGA-266] Add Rafiki python toolkits
      * [SINGA-273] Improve license and contributions
      * [SINGA-284] Add python unittest into Jenkins and link static libs into whl file
      * [SINGA-280] Jenkins CI support
      * [SINGA-288] Publish wheel of PySINGA generated by Jenkins to public servers

  * Documentation and usability
      * [SINGA-263] Create Amazon Machine Image
      * [SINGA-268] Add IPython notebooks to the documentation
      * [SINGA-276] Create docker images
      * [SINGA-289] Update SINGA website automatically using Jenkins
      * [SINGA-295] Add an example of image classification using GoogleNet

  * Bugs fixed
      * [SINGA-245] float as the first operand can not multiply with a tensor object
      * [SINGA-293] Bug from compiling PySINGA on Mac OS X with multiple version of Python

---------------------------------------------------------------
Release Notes - SINGA - Version singa-incubating-1.0.0

SINGA is a general distributed deep learning platform for training big deep learning models over large datasets.

This release includes following features:

  * Core abstractions including Tensor and Device
      * [SINGA-207]  Update Tensor functions for matrices
      * [SINGA-205]  Enable slice and concatenate operations for Tensor objects
      * [SINGA-197]  Add CNMem as a submodule in lib/
      * [SINGA-196]  Rename class Blob to Block
      * [SINGA-194]  Add a Platform singleton
      * [SINGA-175]  Add memory management APIs and implement a subclass using CNMeM
      * [SINGA-173]  OpenCL Implementation
      * [SINGA-171]  Create CppDevice and CudaDevice
      * [SINGA-168]  Implement Cpp Math functions APIs
      * [SINGA-162]  Overview of features for V1.x
      * [SINGA-165]  Add cross-platform timer API to singa
      * [SINGA-167]  Add Tensor Math function APIs
      * [SINGA-166]  light built-in logging for making glog optional
      * [SINGA-164]  Add the base Tensor class


  * IO components for file read/write, network and data pre-processing
      * [SINGA-233]  New communication interface
      * [SINGA-215]  Implement Image Transformation for Image Pre-processing
      * [SINGA-214]  Add LMDBReader and LMDBWriter for LMDB
      * [SINGA-213]  Implement Encoder and Decoder for CSV
      * [SINGA-211]  Add TextFileReader and TextFileWriter for CSV files
      * [SINGA-210]  Enable checkpoint and resume for v1.0
      * [SINGA-208]  Add DataIter base class and a simple implementation
      * [SINGA-203]  Add OpenCV detection for cmake compilation
      * [SINGA-202]  Add reader and writer for binary file
      * [SINGA-200]  Implement Encoder and Decoder for data pre-processing


  * Module components including layer classes, training algorithms and Python binding
      * [SINGA-235]  Unify the engines for cudnn and singa layers
      * [SINGA-230]  OpenCL Convolution layer and Pooling layer
      * [SINGA-222]  Fixed bugs in IO
      * [SINGA-218]  Implementation for RNN CUDNN version
      * [SINGA-204]  Support the training of feed-forward neural nets
      * [SINGA-199]  Implement Python classes for SGD optimizers
      * [SINGA-198]  Change Layer::Setup API to include input Tensor shapes
      * [SINGA-193]  Add Python layers
      * [SINGA-192]  Implement optimization algorithms for SINGA v1 (nesterove, adagrad, rmsprop)
      * [SINGA-191]  Add "autotune" for CudnnConvolution Layer
      * [SINGA-190]  Add prelu layer and flatten layer
      * [SINGA-189]  Generate python outputs of proto files
      * [SINGA-188]  Add Dense layer
      * [SINGA-187]  Add popular parameter initialization methods
      * [SINGA-186]  Create Python Tensor class
      * [SINGA-184]  Add Cross Entropy loss computation
      * [SINGA-183]  Add the base classes for optimizer, constraint and regularizer
      * [SINGA-180]  Add Activation layer and Softmax layer
      * [SINGA-178]  Add Convolution layer and Pooling layer
      * [SINGA-176]  Add loss and metric base classes
      * [SINGA-174]  Add Batch Normalization layer and Local Response Nomalization layer.
      * [SINGA-170]  Add Dropout layer and CudnnDropout layer.
      * [SINGA-169]  Add base Layer class for V1.0


  * Examples
      * [SINGA-232]  Alexnet on Imagenet
      * [SINGA-231]  Batchnormlized VGG model for cifar-10
      * [SINGA-228]  Add Cpp Version of Convolution and Pooling layer
      * [SINGA-227]  Add Split and Merge Layer and add ResNet Implementation

  * Documentation
      * [SINGA-239]  Transfer documentation files of v0.3.0 to github
      * [SINGA-238]  RBM on mnist
      * [SINGA-225]  Documentation for installation and Cifar10 example
      * [SINGA-223]  Use Sphinx to create the website

  * Tools for compilation and some utility code
      * [SINGA-229]  Complete install targets
      * [SINGA-221]  Support for Travis-CI
      * [SINGA-217]  build python package with setup.py
      * [SINGA-216]  add jenkins for CI support
      * [SINGA-212]  Disable the compilation of libcnmem if USE_CUDA is OFF
      * [SINGA-195]  Channel for sending training statistics
      * [SINGA-185]  Add CBLAS and GLOG detection for singav1
      * [SINGA-181]  Add NVCC supporting for .cu files
      * [SINGA-177]  Add fully cmake supporting for the compilation of singa_v1
      * [SINGA-172]  Add CMake supporting for Cuda and Cudnn libs

----------------------------------------------------------
Release Notes - SINGA - Version singa-incubating-0.3.0

SINGA is a general distributed deep learning platform for training big deep learning models over large datasets.

This release includes following features:

  * GPU Support
    * [SINGA-131] Implement and optimize hybrid training using both CPU and GPU
    * [SINGA-136] Support cuDNN v4
    * [SINGA-134] Extend SINGA to run over a GPU cluster
    * [SINGA-157] Change the priority of cudnn library and install libsingagpu.so

  * Remove Dependencies
    * [SINGA-156] Remove the dependency on ZMQ for single process training
    * [SINGA-155] Remove zookeeper for single-process training

  * Python Binding
    * [SINGA-126] Python Binding for Interactive Training

  * Other Improvements
    * [SINGA-80] New Blob Level and Address Level Math Operation Interface
    * [SINGA-130] Data Prefetching
    * [SINGA-145] New SGD based optimization Updaters: AdaDelta, Adam, AdamMax

  * Bugs Fixed
    * [SINGA-148] Race condition between Worker threads and Driver
    * [SINGA-150] Mesos Docker container failed
    * [SIGNA-141] Undesired Hash collision when locating process id to worker…
    * [SINGA-149] Docker build fail
    * [SINGA-143] The compilation cannot detect libsingagpu.so file


-----------------------------------------
Release Notes - SINGA - Version singa-incubating-0.2.0

SINGA is a general distributed deep learning platform for training big deep learning models over large datasets. It is
designed with an intuitive programming model based on the layer abstraction. SINGA supports a wide variety of popular
deep learning models.

This release includes following features:

  * Programming model
    * [SINGA-80] New Blob Level and Address Level Math Operation Interface
    * [SINGA-82] Refactor input layers using data store abstraction
    * [SINGA-87] Replace exclude field to include field for layer configuration
    * [SINGA-110] Add Layer member datavec_ and gradvec_
    * [SINGA-120] Implemented GRU and BPTT (BPTTWorker)

  * Neuralnet layers
    * [SINGA-91] Add SoftmaxLayer and ArgSortLayer
    * [SINGA-106] Add dummy layer for test purpose
    * [SINGA-120] Implemented GRU and BPTT (GRULayer and OneHotLayer)

  * GPU training support
    * [SINGA-100] Implement layers using CUDNN for GPU training
    * [SINGA-104] Add Context Class
    * [SINGA-105] Update GUN make files for compiling cuda related code
    * [SINGA-98] Add Support for AlexNet ImageNet Classification Model


  * Model/Hybrid partition
    * [SINGA-109] Refine bridge layers
    * [SINGA-111] Add slice, concate and split layers
    * [SINGA-113] Model/Hybrid Partition Support

  * Python binding
    * [SINGA-108] Add Python wrapper to singa

  * Predict-only mode
    * [SINGA-85] Add functions for extracting features and test new data

  * Integrate with third-party tools
    * [SINGA-11] Start SINGA on Apache Mesos
    * [SINGA-78] Use Doxygen to generate documentation
    * [SINGA-89] Add Docker support

  * Unit test
    * [SINGA-95] Add make test after building

  * Other improvment
    * [SINGA-84] Header Files Rearrange
    * [SINGA-93] Remove the asterisk in the log tcp://169.254.12.152:*:49152
    * [SINGA-94] Move call to google::InitGoogleLogging() from Driver::Init() to main()
    * [SINGA-96] Add Momentum to Cifar10 Example
    * [SINGA-101] Add ll (ls -l) command in .bashrc file when using docker
    * [SINGA-114] Remove short logs in tmp directory
    * [SINGA-115] Print layer debug information in the neural net graph file
    * [SINGA-118] Make protobuf LayerType field id easy to assign
    * [SIGNA-97] Add HDFS Store

  * Bugs fixed
    * [SINGA-85] Fix compilation errors in examples
    * [SINGA-90] Miscellaneous trivial bug fixes
    * [SINGA-107] Error from loading pre-trained params for training stacked RBMs
    * [SINGA-116] Fix a bug in InnerProductLayer caused by weight matrix sharing

-----------------------------------------

Features included in singa-incubating-0.1.0:

  * Job management
    * [SINGA-3] Use Zookeeper to check stopping (finish) time of the system
    * [SINGA-16] Runtime Process id Management
    * [SINGA-25] Setup glog output path
    * [SINGA-26] Run distributed training in a single command
    * [SINGA-30] Enhance easy-to-use feature and support concurrent jobs
    * [SINGA-33] Automatically launch a number of processes in the cluster
    * [SINGA-34] Support external zookeeper service
    * [SINGA-38] Support concurrent jobs
    * [SINGA-39] Avoid ssh in scripts for single node environment
    * [SINGA-43] Remove Job-related output from workspace
    * [SINGA-56] No automatic launching of zookeeper service
    * [SINGA-73] Refine the selection of available hosts from host list

  * Installation with GNU Auto tool
    * [SINGA-4] Refine thirdparty-dependency installation
    * [SINGA-13] Separate intermediate files of compilation from source files
    * [SINGA-17] Add root permission within thirdparty/install.
    * [SINGA-27] Generate python modules for proto objects
    * [SINGA-53] Add lmdb compiling options
    * [SINGA-62] Remove building scrips and auxiliary files
    * [SINGA-67] Add singatest into build targets

  * Distributed training
    * [SINGA-7] Implement shared memory Hogwild algorithm
    * [SINGA-8] Implement distributed Hogwild
    * [SINGA-19] Slice large Param objects for load-balance
    * [SINGA-29] Update NeuralNet class to enable layer partition type customization
    * [SINGA-24] Implement Downpour training framework
    * [SINGA-32] Implement AllReduce training framework
    * [SINGA-57] Improve Distributed Hogwild

  * Training algorithms for different model categories
    * [SINGA-9] Add Support for Restricted Boltzman Machine (RBM) model
    * [SINGA-10] Add Support for Recurrent Neural Networks (RNN)

  * Checkpoint and restore
    * [SINGA-12] Support Checkpoint and Restore

  * Unit test
    * [SINGA-64] Add the test module for utils/common

  * Programming model
    * [SINGA-36] Refactor job configuration, driver program and scripts
    * [SINGA-37] Enable users to set parameter sharing in model configuration
    * [SINGA-54] Refactor job configuration to move fields in ModelProto out
    * [SINGA-55] Refactor main.cc and singa.h
    * [SINGA-61] Support user defined classes
    * [SINGA-65] Add an example of writing user-defined layers

  * Other features
    * [SINGA-6] Implement thread-safe singleton
    * [SINGA-18] Update API for displaying performance metric
    * [SINGA-77] Integrate with Apache RAT

Some bugs are fixed during the development of this release
  * [SINGA-2] Check failed: zsock_connect
  * [SINGA-5] Server early terminate when zookeeper singa folder is not initially empty
  * [SINGA-15] Fixg a bug from ConnectStub function which gets stuck for connecting layer_dealer_
  * [SINGA-22] Cannot find openblas library when it is installed in default path
  * [SINGA-23] Libtool version mismatch error.
  * [SINGA-28] Fix a bug from topology sort of Graph
  * [SINGA-42] Issue when loading checkpoints
  * [SINGA-44] A bug when reseting metric values
  * [SINGA-46] Fix a bug in updater.cc to scale the gradients
  * [SINGA-47] Fix a bug in data layers that leads to out-of-memory when group size is too large
  * [SINGA-48] Fix a bug in trainer.cc that assigns the same NeuralNet instance to workers from diff groups
  * [SINGA-49] Fix a bug in HandlePutMsg func that sets param fields to invalid values
  * [SINGA-66] Fix bugs in Worker::RunOneBatch function and ClusterProto
  * [SINGA-79] Fix bug in singatool that can not parse -conf flag

Features planned for the next release
  * [SINGA-11] Start SINGA using Mesos
  * [SINGA-31] Extend Blob to support xpu (cpu or gpu)
  * [SINGA-35] Add random number generators
  * [SINGA-40] Support sparse Param update
  * [SINGA-41] Support single node single GPU training