AttributeError: 'module' object has no attribute 'img_rnorm_layer' - while running image_classification demo training about paddle

[qingqing01]

You just need to remove img_rnorm_layer in line 39 in python/paddle/trainer_config_helpers/layers.py. We have fixed it and will merge it soon.

[deepcodingman]

Thanks. that problem got resolved but now I am getting another error.

[18:40 ~/paddle/src/paddle/demo/image_classification] > ./train.sh
I0902 18:40:26.984419 27728 Util.cpp:144] commandline: /home/deepbeast/paddle/bin/bin/../opt/ paddle/bin/paddle_trainer --config=vgg_16_cifar.py --dot_period=10 --log_period=100 --test_al l_data_in_one_period=1 --use_gpu=1 --trainer_count=1 --num_passes=200 --save_dir=./cifar_vgg_ model
I0902 18:40:29.325484 27728 Util.cpp:113] Calling runInitFunctions
I0902 18:40:29.325641 27728 Util.cpp:126] Call runInitFunctions done.
[INFO 2016-09-02 18:40:29,394 layers.py:1435] channels=3 size=3072
[INFO 2016-09-02 18:40:29,394 layers.py:1435] output size for conv_0 is 32
[INFO 2016-09-02 18:40:29,397 layers.py:1435] channels=64 size=65536
[INFO 2016-09-02 18:40:29,398 layers.py:1435] output size for conv_1 is 32
[INFO 2016-09-02 18:40:29,400 layers.py:1496] output size for pool_0 is 16_16
[INFO 2016-09-02 18:40:29,401 layers.py:1435] channels=64 size=16384
[INFO 2016-09-02 18:40:29,401 layers.py:1435] output size for conv_2 is 16
[INFO 2016-09-02 18:40:29,404 layers.py:1435] channels=128 size=32768
[INFO 2016-09-02 18:40:29,404 layers.py:1435] output size for conv_3 is 16
[INFO 2016-09-02 18:40:29,406 layers.py:1496] output size for pool_1 is 8_8
[INFO 2016-09-02 18:40:29,407 layers.py:1435] channels=128 size=8192
[INFO 2016-09-02 18:40:29,408 layers.py:1435] output size for conv_4 is 8
[INFO 2016-09-02 18:40:29,410 layers.py:1435] channels=256 size=16384
[INFO 2016-09-02 18:40:29,410 layers.py:1435] output size for conv_5 is 8
[INFO 2016-09-02 18:40:29,413 layers.py:1435] channels=256 size=16384
[INFO 2016-09-02 18:40:29,413 layers.py:1435] output size for conv_6 is 8
[INFO 2016-09-02 18:40:29,415 layers.py:1496] output size for pool_2 is 4_4
[INFO 2016-09-02 18:40:29,416 layers.py:1435] channels=256 size=4096
[INFO 2016-09-02 18:40:29,417 layers.py:1435] output size for conv_7 is 4
[INFO 2016-09-02 18:40:29,419 layers.py:1435] channels=512 size=8192
[INFO 2016-09-02 18:40:29,419 layers.py:1435] output size for conv_8 is 4
[INFO 2016-09-02 18:40:29,421 layers.py:1435] channels=512 size=8192
[INFO 2016-09-02 18:40:29,422 layers.py:1435] output size for conv_9 is 4
[INFO 2016-09-02 18:40:29,424 layers.py:1496] output size for pool_3 is 2_2
[INFO 2016-09-02 18:40:29,424 layers.py:1496] output size for pool_4 is 1_1
[INFO 2016-09-02 18:40:29,428 networks.py:1122] The input order is [image, label]
[INFO 2016-09-02 18:40:29,429 networks.py:1125] The output order is [cost_0]
I0902 18:40:29.442795 27728 Trainer.cpp:169] trainer mode: Normal
F0902 18:40:29.443608 27728 hl_gpu_matrix_kernel.cuh:181] Check failed: cudaSuccess == err (0 vs. 8) [hl_gpu_apply_unary_op failed] CUDA error: invalid device function
*_* Check failure stack trace: ***
@ 0x7feafca9f5cd google::LogMessage::Fail()
@ 0x7feafcaa1433 google::LogMessage::SendToLog()
@ 0x7feafca9f15b google::LogMessage::Flush()
@ 0x7feafcaa1e1e google::LogMessageFatal::~LogMessageFatal()
@ 0x7ed69c hl_gpu_apply_unary_op<>()
@ 0x7b3c3b paddle::BaseMatrixT<>::applyUnary<>()
@ 0x7b37c5 paddle::BaseMatrixT<>::applyUnary<>()
@ 0x7952ec paddle::BaseMatrixT<>::zero()
@ 0x65e3a0 paddle::Parameter::enableType()
@ 0x66e001 paddle::parameterInitNN()
@ 0x671ca5 paddle::NeuralNetwork::init()
@ 0x66797b paddle::GradientMachine::create()
@ 0x6f1cfd paddle::TrainerInternal::init()
@ 0x6eb38b paddle::Trainer::init()
@ 0x547e65 main
@ 0x7feafbc29830 __libc_start_main
@ 0x5501a9 _start
@ (nil) (unknown)
/home/deepbeast/paddle/bin/bin/paddle: line 46: 27728 Aborted (core dumped) $ {DEBUGGER} $MYDIR/../opt/paddle/bin/paddle_trainer ${@:2}
No data to plot. Exiting!

[emailweixu]

Can you provide your GPU card type and CUDA version? Thanks

[deepcodingman]

Please see below:

[19:08 ~] > lspci -nn | grep '[03'
01:00.0 VGA compatible controller [0300]: NVIDIA Corporation GP104 [GeForce GTX 1080] [10de:1b80](rev a1)
02:00.0 VGA compatible controller [0300]: NVIDIA Corporation GP104 [GeForce GTX 1080] [10de:1b80](rev a1)

[19:08 ~] > cat /usr/local/cuda/version.txt
CUDA Version 8.0.27
[19:08 ~] >

[gangliao]

Hi, invalid device function indicates that you have a CUDA / GPU incompatibility.
Since you use GTX 1080, you can modify CMake to fix it.

1. open cmake/flags.cmake and add following code:

if (CUDA_VERSION VERSION_GREATER "8.0")

list(APPEND __arch_flags " -gencode arch=compute_60,code=sm_60")

endif()

1. then, rebuild the project

[deepcodingman]

Now I am getting following error during running 'make -j all'

[ 16%] Built target paddle_utils
[ 16%] Building NVCC (Device) object paddle/gserver/CMakeFiles/paddle_gserver.dir/layers/paddle_gserver_generated_LstmCompute.cu.o
[ 19%] Built target paddle_pserver
[ 19%] Built target paddle_network
[ 19%] Building NVCC (Device) object paddle/gserver/CMakeFiles/paddle_gserver.dir/layers/paddle_gserver_generated_GruCompute.cu.o
[ 23%] Built target paddle_api
[ 26%] Built target paddle_trainer_lib
[ 31%] Built target paddle_parameter
/usr/local/cuda/include/surface_functions.h(134): error: expected a ";"

/usr/local/cuda/include/surface_functions.h(135): error: expected a ";"

/usr/local/cuda/include/surface_functions.h(136): error: expected a ";"

[gangliao]

Basically, we do not test PaddlePaddle on cuda 8.0. There maybe exist some incompatibility.
You can use make clean; cmake ..; make VERBOSE=1 to figure out where the problem comes from.

[deepcodingman]

Thanks. I will give it a try again later once there is support for CUDA 8.0

[hedaoyuan]

CUDA 8.0 has support.