Caffe: libcudart.so.8.0: cannot open shared object file: No such file or directory

Created on 2 Nov 2016 · 11Comments · Source: BVLC/caffe

.......
[==========] 2081 tests from 277 test cases ran. (324463 ms total)
[ PASSED ] 2081 tests.

i have make runtest,as it is showed above,the runtest have passed.But when i try to run mnist example,the error happened as below:

``:~/caffe$ sudo sh ./examples/mnist/create_mnist.sh
Creating lmdb...
build/examples/mnist/convert_mnist_data.bin: error while loading shared libraries: libcudart.so.8.0: cannot open shared object file: No such file or directory

many people said i should add the path in the bashrc document,and i have added as below:

export LD_LIBRARY_PATH=/usr/local/cuda-8.0/lib64:$LD_LIBRARY_PATH
export PATH=/usr/local/cuda-8.0/bin:$PATH
export PATH

but i try again,the error happened again.i do not know what i did wrong,if anyone know,please tell me what i should do,thanks in advance.

Source

ghost

Most helpful comment

@Edwardmark

As @fujihuang mentioned, just add the path
export LD_LIBRARY_PATH=/usr/local/cuda-8.0/lib64:$LD_LIBRARY_PATH
export PATH=/usr/local/cuda-8.0/bin:$PATH

However, fujihuang adds the third comment which makes his PATH to be empty.

zhenni on 23 May 2017

👍54 ❤8

All 11 comments

Hard to understand why is this one closed ? the issue still exists.

.build_release/tools/caffe
.build_release/tools/caffe: error while loading shared libraries: libopenblas.so.0: cannot open shared object file: No such file or directory
Makefile:533: recipe for target 'runtest' failed
make: *** [runtest] Error 127

Shubhamcl on 10 Mar 2017

I met the same error, please help me

Edwardmark on 14 May 2017

@Edwardmark

As @fujihuang mentioned, just add the path
export LD_LIBRARY_PATH=/usr/local/cuda-8.0/lib64:$LD_LIBRARY_PATH
export PATH=/usr/local/cuda-8.0/bin:$PATH

However, fujihuang adds the third comment which makes his PATH to be empty.

zhenni on 23 May 2017

👍54 ❤8

Thank you @zhenni . It helped

martianvenusian on 8 Aug 2017

thank you @zhenni. You saved my day

martianvenusian on 15 Sep 2017

hi @zhenni,

your solution does not work. do you have other suggestion how to solve it?

zimenglan-sysu-512 on 6 Dec 2017

sudo ldconfig /usr/local/cuda/lib64
Fine for me.
Please refer this link for more information.
https://github.com/tensorflow/tensorflow/issues/5343

SelenaKong on 14 Mar 2018

👍29

if the first method mentioned above can't help,maybe you can try this,
go into the dir /usr/local/cuda/include/host_config.h, then comments"#error-- unsupported GNU version! gcc versions later than 4.9 are not supported!"
like this "//#error-- unsupported GNU version! gcc versions later than 4.9 are not supported!"

calf1in1cup on 5 Apr 2018

👍3

@ghost, I would like to bring some more information to resolve the error as i was also facing the same issue. First of all as zhenni said you need to add the path in the file .bashrc (ubuntu user). For example see the below line I have added in my .bashrc file.

LD_LIBRARY_PATH=/usr/local/lib
export LD_LIBRARY_PATH=/usr/local/cuda-10.0/lib64:${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}

Please do add one more point once you added these files in the .bashrc it is important to reflect the changes in the file using the command .bashrc. Now, restart the terminal to see the change. Please run the below command in the terminal

echo $LD_LIBRARY_PATH

You will see the output like this

/usr/local/cuda-10.0/lib64::/usr/local/lib
Once, you are successfully in running these steps successfully, please go to the following path

cd $HOME/cudnn_samples_v7/mnistCUDNN

Please run the below command

make clean && make

This command is very important before running the command ./mnistCUDNN.

You will get the message as shown below

cuddn_error_resolve