Documentation

• Documentation
  • Installation
    • From Conda
    • From source
      • Use Conda to build SINGA
      • Use native tools to build SINGA on Ubuntu
      • Compile SINGA on Windows
      • More details about the compilation options
      • USE_MODULES
        • USE_PYTHON
        • USE_CUDA
        • USE_OPENCL
        • USE_MKLDNN
        • PACKAGE
    • FAQ
  • Software Stack
    • Core
    • Model
    • IO
  • Benchmark for Distributed training
  • Device
    • Specific devices
    • Python API
    • CPP API
  • Tensor
    • Tensor implementation
    • Python API
    • CPP API
  • Layer
    • Python API
    • CPP API
  • FeedForward Net
  • Initializer
    • Python API
    • CPP API
  • Loss
  • Metric
  • Optimizer
  • Autograd in Singa
    • Relevant Modules
      • Tensor
      • Operation
      • Layer
    • Examples
      • Operation only
        • Import packages
        • Create weight matrix and bias vector
        • Training
      • Operation + Layer
        • Create the layers
        • Define the forward function
        • Training
  • Data
  • Image Tool
  • Snapshot
  • Caffe Converter
  • Utils
  • Model Zoo
    • Train CNN over Cifar-10
      • Instructions
        • SINGA installation
        • Data preparation
        • Training
        • Prediction
    • Train Char-RNN over plain text
      • Instructions
    • Train AlexNet over ImageNet
      • Instructions
        • Compile SINGA
        • Data download
        • Data preprocessing
        • Training
    • name: GoogleNet on ImageNet SINGA version: 1.0.1 SINGA commit: 8c990f7da2de220e8a012c6a8ecc897dc7532744 parameter_url: https://s3-ap-southeast-1.amazonaws.com/dlfile/bvlc_googlenet.tar.gz parameter_sha1: 0a88e8948b1abca3badfd8d090d6be03f8d7655d license: unrestricted https://github.com/BVLC/caffe/tree/master/models/bvlc_googlenet
    • Image Classification using GoogleNet
      • Instructions
      • Details