Skip to content
/ ConvNN Public

A simple CNN training framework support on CPU and GPU(CUDNN)

Notifications You must be signed in to change notification settings

gty111/ConvNN

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

17 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

A CNN training framework

just for play

Features

  • An example to use CUDNN

  • A simple CNN training framework without any existing machine learning framework

  • You can build CNN like building blocks through this framework

  • I provide two examples(One is on MNIST(achieve 98% Acc),the other is on CIFAR-10(achieve 70% Acc)) to build CNN through this framework

  • whose backend is on GPU (call CUDNN to accelerate)

Doc

You can find document about this repo from here.

Env

  • CUDA
  • CUDNN
  • CUBLAS
  • GPU support CUDA and CUDNN

before run anything

export CUDNN=<your path to CUDNN>

Prepare data

make data

Run example

At each run, the standard output will be recorded in the log file

MNIST

you can find log (which I run) at here

make mnist

CIFAR

you can find log (which I run) at here

make cifar

Run test

test cudnn

make example 
make test_conv 
make test_fc 
make test_max_pool 
make test_relu

test layer

make fc
make act 
make pool 
make softmax 
make batchnorm

About

A simple CNN training framework support on CPU and GPU(CUDNN)

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published