hijkzzz / cuda-neural-network

Convolutional Neural Network with CUDA (MNIST 99.23%)

🔭 I'm a Coding Lover.

cuda-neural-network's Issues

nvcc fatal : redefinition of argument 'optimize'

I got an issue when I compile the project using make , here is the output:

➜ build make

Scanning dependencies of target cu

[ 5%] Building CUDA object CMakeFiles/cu.dir/src/cuda/blas.cu.o

nvcc fatal : redefinition of argument 'optimize'

CMakeFiles/cu.dir/build.make:62: recipe for target 'CMakeFiles/cu.dir/src/cuda/blas.cu.o' failed

make[2]: *** [CMakeFiles/cu.dir/src/cuda/blas.cu.o] Error 1

CMakeFiles/Makefile2:67: recipe for target 'CMakeFiles/cu.dir/all' failed

make[1]: *** [CMakeFiles/cu.dir/all] Error 2

Makefile:83: recipe for target 'all' failed

make: *** [all] Error 2

It seems like a nvcc compiling issue using optimizing arguments like '-O3' or '-O2', but I just can not figure out. Can someone help me? Thx!

loss: nan

The loss shows nan when I train the code, is there any solution?

卷积的反向传播

src/cuda/conv.cu 的第120行，计算height_col的公式是“int height_col = height + 2 * pad_h - kernel_h / stride_h + 1;”
为什么不是 “int height_col = (height + 2 * pad_h - kernel_h )/ stride_h + 1;” 是不是忘写了个括号呀

build error: redefinition of argument 'optimize'

Here is what cmake found:

-- The CUDA compiler identification is NVIDIA 10.2.89
-- The CXX compiler identification is GNU 8.4.0
-- Check for working CUDA compiler: /opt/cuda/bin/nvcc
-- Check for working CUDA compiler: /opt/cuda/bin/nvcc - works
-- Detecting CUDA compiler ABI info
-- Detecting CUDA compiler ABI info - done
-- Detecting CUDA compile features
-- Detecting CUDA compile features - done
-- Check for working CXX compiler: /opt/cuda/bin/g++
-- Check for working CXX compiler: /opt/cuda/bin/g++ - works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Looking for C++ include pthread.h
-- Looking for C++ include pthread.h - found
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Failed
-- Looking for pthread_create in pthreads
-- Looking for pthread_create in pthreads - not found
-- Looking for pthread_create in pthread
-- Looking for pthread_create in pthread - found
-- Found Threads: TRUE  
-- Found CUDA: /opt/cuda (found version "10.2") 
-- Configuring done
-- Generating done
-- Build files have been written to: /home/user/test/cuda-neural-network/build

I'm getting the following error:

Scanning dependencies of target cu
[  6%] Building CUDA object CMakeFiles/cu.dir/src/cuda/blas.cu.o
nvcc fatal   : redefinition of argument 'optimize'
make[2]: *** [CMakeFiles/cu.dir/build.make:83: CMakeFiles/cu.dir/src/cuda/blas.cu.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:98: CMakeFiles/cu.dir/all] Error 2
make: *** [Makefile:104: all] Error 2

Any ideas?

loss -nan

the file can be compiled but the loss is constantly -nan, can you plz show me a clue on where the dataset should be placed?

Any plans for save model and real image inference?

Thanks for your awesome code share!

Recently I compare some cuda and cpp mnist train tools from github, this repo is almost the fastest and most accurate one. With functions of saving model and real 28*28 image inference, it is not only a great research tool but also a fast and accurate app.

Why am I getting "nvcc fatal : redefinition of argument 'optimize'"?

Hi, thanks for your project. I am trying to learn cuda with the help of your code. When I tried to install and make this cuda project, I got the error "nvcc fatal : redefinition of argument 'optimize". It seems that it is a bug from nvcc? May you help me?

The version number of the cub library that corresponds to this cuda version.

No kernel image is available for execution on the device

Hi.
I tried this program on the server but failed. Coud you help me figure out what's wrong here?
Log is below: