-
Notifications
You must be signed in to change notification settings - Fork 754
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
cuda 8.0 | cudnn 6.0 | nnconv_cudnn.cu: error: too few arguments in function call #937
Comments
Anyone has any idea how to fix this? Or which version should I switch to? |
Hi, we haven't implemented the cudnn 6.0 support yet, as it has been
released few days back...
But it should work with cudnn 5.1 just fine...
…On 30 March 2017 at 14:54, xwguo ***@***.***> wrote:
Anyone has any idea how to fix this? Or which version should I switch to?
I would really appreciate your help!
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#937 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/ABe3FYUiM0ilnou4b8ULW8JOc9T-8jFsks5rq7QkgaJpZM4MuY-3>
.
|
@lenck Thanks a lot for your prompt reply. It compiles with 5.1! I did add echo $LD_LIBRARY_PATH^ shows Error occurred in nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value1,oh=value1,ow=value1,multiple_grids=value1) and it did not run to completion. Thanks a lot!!! |
I ran across the same problem, also with using ubuntu 16.04, matlab 2017a, cudnn 6.0, GTX 1070, cuda 8.0, g++/gcc 5.4. I did try downgrading to cudnn 5.1 and g++/gxx 4.9, which didn't change anything. I found a solution here: https://groups.google.com/forum/#!topic/caffe-users/aoJdbIxT_PE The suggestion was to have the needed libraries preload: |
Thank @unvisualism ! I had the same problem with ubuntu 16.04, matlab 2017a, g++/gcc 5.4 and without GPU. I add the following content in ~/.bashrc and fixed the problem. export LD_PRELOAD=$LD_PRELOAD:/usr/lib/x86_64-linux-gnu/libstdc++.so.6:/usr/lib/x86_64-linux-gnu/libprotobuf.so.9 By the way, don't forget to execute |
Hi I wonder if cudnn 6.0 support have been implemented or not. |
Thanks @daa233. This works for me! |
Hi, the cudnn v6.0 and v7.0 support will be in the next release which should be coming out in few days... Sorry that it took so long. There were just few minor changes needed to fix it... |
Hi, I'm getting the same error on Windows. How can I preload those libraries on Windows? |
I don’t know how you would do it on Windows, I’m sorry.
… On Oct 8, 2017, at 5:11 PM, varunsaboo ***@***.***> wrote:
Hi, I'm getting the same error on Windows. How can I preload those libraries on Windows?
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub, or mute the thread.
|
I have the same error on Ubuntu16.04, GCC5.5. I also try @daa233 's solution, but it doesn't work for me.
|
I am using ubuntu 16.04, matlab 2017a, cudnn 6.0, GTX 1070, cuda 8.0, g++/gcc 5.4.
I could do the vl_compilenn and vl_compilenn with gpu but without cudnn support.
The only error I got is from vl_compilenn with cudnn enable, as shown in below.
Besides, I could compile for all three ways with cudnn-rc4, but when it comes to vl_testnn, there would be lots of missing symbols.
error message with vl_compilenn enable cudnn-6:
/home/emily/Documents/FYP/fayao_3_30/libs/matconvnet-1.0-beta24/matlab/src/bits/impl/nnconv_cudnn.cu(137): error: too few arguments in function call
/home/emily/Documents/FYP/fayao_3_30/libs/matconvnet-1.0-beta24/matlab/src/bits/impl/nnconv_cudnn.cu(356): error: too few arguments in function call
2 errors detected in the compilation of "/tmp/tmpxft_000076fb_00000000-7_nnconv_cudnn.cpp1.ii".
Error using vl_compilenn>nvcc_compile (line 540)
Command "/usr/local/cuda-8.0/bin/nvcc" -c
"/home/emily/Documents/FYP/fayao_3_30/libs/matconvnet-1.0-beta24/matlab/src/bits/impl/nnconv_cudnn.cu"
-DNDEBUG -DENABLE_GPU -DENABLE_CUDNN -I"/usr/local/cuda-8.0/cudnn-6/include" -DENABLE_DOUBLE
-gencode=arch=compute_61,code="sm_61,compute_61" -I"/home/emily/Documents/FYP/R2017a/extern/include"
-I"/home/emily/Documents/FYP/R2017a/toolbox/distcomp/gpu/extern/include" -Xcompiler
-mssse3,-ftree-vect-loop-version,-ffast-math,-funroll-all-loops
-gencode=arch=compute_61,code="sm_61,compute_61" -O3 -Xcompiler -fPIC -D_FORCE_INLINES -o
"/home/emily/Documents/FYP/fayao_3_30/libs/matconvnet-1.0-beta24/matlab/mex/.build/bits/impl/nnconv_cudnn.o"
failed.
Error in vl_compilenn (line 485)
nvcc_compile(opts, srcs{i}, objfile, flags.nvcc)
error message with cudnn-rc4 while doing vl_testnn:
Error occurred in nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value1,oh=value1,ow=value1,multiple_grids=value1) and it did not run to completion.
---------
Error ID:
---------
'MATLAB:mex:ErrInvalidMEXFile'
--------------
Error Details:
--------------
Invalid MEX-fileError occurred in nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value1,oh=value1,ow=value1,multiple_grids=value1) and it did not run to completion.
The text was updated successfully, but these errors were encountered: