Easy installation
Easy installation using Conda, Pip, Docker and from Source
Database integration
Models are trained with SINGA and can be queried in the RDBMS
Model zoo
Various example deep learning models are provided in SINGA repo on Github and on Google Colab
Distributed training
SINGA supports data parallel training across multiple GPUs (on a single node or across different nodes)
Automatic gradient calculation
SINGA records the computation graph and applies the backward propagation automatically after forward propagation
Memory optimization
The optimization of memory are implemented in the Device class
Parameter optimization
SINGA supports various popular optimizers including stochastic gradient descent with momentum, Adam, RMSProp, and AdaGrad, etc
Interoperability
SINGA supports loading ONNX format models and saving models defined using SINGA APIs into ONNX format, which enables AI developers to use models across different libraries and tools
Time profiling
SINGA supports the time profiling of each of the operators buffered in the graph
SINGA has a well architected software stack and easy-to-use Python interface to improve usability
SINGA trains the deep learning models which can be queried as stored procedures of the RDBMS
SINGA parallelizes the training and optimizes the communication cost to improve training scalability
SINGA builds a computational graph to optimize the training speed and memory footprint
Users of Apache SINGA
Apache SINGA powers the following organizations and companies...