Comments (12)
Hi Ken,
Which version of openmpi are you using?
You can try the following steps:
cd ~
wget http://www.open-mpi.org/software/ompi/v1.10/downloads/openmpi-1.10.1.tar.gz
gunzip -c openmpi-1.10.1.tar.gz | tar xf -
cd openmpi-1.10.1
./configure --prefix=$HOME/mpi
make -j all install
export PATH=$HOME/mpi/bin:$PATH
[Change directory to CNTK repository root, e.g. $HOME/src/CNTK/]
mkdir -p build/release && cd build/release && ../../configure --with-buildtype=release && make -j all
cd .. && mkdir debug && cd debug && ../../configure --with-buildtype=debug && make -j all
export PATH=$PATH:$HOME/src/CNTK/build/debug/bin/
Let us know.
Thanks,
Gaizka
from cntk.
Also I spotted you use Cuda 7.5, while the code expects 7.0, so there might be extra errors, after this one is resolved with such config
from cntk.
Hi Gaizka,
I did use the openmpi version 1.10.1 and installed it into my home folder as you suggested.
I also tried building as debug as you recommended, and got the same error message.
BR
Ken
from cntk.
Hi @alexeyo26
I am a little uncomfortable installing CUDA kit v7.0 since ver7.5 is working fine with other tools. Do you know if it is safe to install two different versions of the CUDA tool kits?
BTW, when will Microsoft release a binary pkg with CUDA 7.5 for Linux?
Thanks
Ken
from cntk.
Type 'which mpic++' and ensure it's pointing to the bin folder of your newly built OpenMPI. Make sure your path is set properly with the path to the bin folder of the newer OpenMPI ahead of the path to any other instance on your system so that the correct mpic++ wrapper is called in the build process.
from cntk.
Hi Ross,
Thank you for the tip! Yeah, it was my mistake. I had my $HOME/mpi/bin
appended to $PATH, instead of prefixing it.
It works now.
BR
Ken
On 01/28/2016 07:34 PM, Ross Wightman wrote:
Type 'which mpic++' and ensure it's pointing to the bin folder of your
newly built OpenMPI. Make sure your path is set properly with the path
to the bin folder of the newer OpenMPI ahead of the path to any other
instance on your system so that the correct mpic++ wrapper is called
in the build process.—
Reply to this email directly or view it on GitHub
#48 (comment).
from cntk.
@ktang2k Hi,Ken, I also met the same problem ,which was the 'Source/SGDLib/SimpleDistGradAggregator.h:282:186: error: ‘MPI_Iallreduce’ was not declared in this scope'. I have carefullly checked the path of mpic++, I am sorry ,but I don't understand what did you do. Could you give me more suggestions? Thank you. by the way, System :ubuntu15.04 ,cuda Toolkit v7.5
from cntk.
Hi,
In my case, I had to make sure the correct version of MPI is called by the
builder, by prefixing its lib and bin folders to PATH and LD_LIBRARY_PATH
Ken
On Mar 4, 2016 5:21 AM, "linquanxu" [email protected] wrote:
@ktang2k https://github.com/ktang2k Hi,Ken, I also met the same problem
,which was the 'Source/SGDLib/SimpleDistGradAggregator.h:282:186: error:
‘MPI_Iallreduce’ was not declared in this scope'. I have carefullly checked
the path of mpic++, I am sorry ,but I don't understand what did you do.
Could you give me more suggestions? Thank you.—
Reply to this email directly or view it on GitHub
#48 (comment).
from cntk.
@ktang2k Hi Ken,Thank you for your reply,My config file (/etc/bash.bashrc) as follows:export PATH=/usr/local/mpi/bin:/usr/local/MATLAB/R2013a/bin:/usr/local/cuda-7.0/bin:$PATH
,
export LD_LIBRARY_PATH=/usr/local/mpi/lib:/opt/acml5.3.1/ifort64/lib:/opt/acml5.3.1/ifort64_mp/lib:/usr/local/cudnn-4.0/cuda/lib64:/usr/local/cuda-7.0/lib64:$LD_LIBRARY_PATH
. but I got the same error message.How can I solve the problem?
from cntk.
Hi linquanxu, I could not see why your build would fail.
The only diff I spot is that you are a using system installed mpi. I am
using locally installed (to my home folder) as suggested by the setup
instruction. But I seriously doubt that could be the issue :)
Sorry, I am out of ideas now.
Ken
On Sun, Mar 6, 2016 at 9:05 AM, linquanxu [email protected] wrote:
@ktang2k https://github.com/ktang2k Hi Ken,Thank you for your reply,My
config file (/etc/bash.bashrc) as follows:export
PATH=/usr/local/mpi/bin:/usr/local/MATLAB/R2013a/bin:/usr/local/cuda-7.0/bin:$PATH
,export
LD_LIBRARY_PATH=/usr/local/mpi/lib:/opt/acml5.3.1/ifort64/lib:/opt/acml5.3.1/ifort64_mp/lib:/usr/local/cudnn-4.0/cuda/lib64:/usr/local/cuda-7.0/lib64:$LD_LIBRARY_PATH.
but I got the same error message.How can I solve the problem?—
Reply to this email directly or view it on GitHub
#48 (comment).
from cntk.
@ktang2k Hi,Key,Thanks again,But issue still exists.I should have check this more clearly.
from cntk.
Do you solve this issue @linquanxu ? I use
OS: ubuntu14.04.4
cude: cuda 7.5.18
openmpi: openmpi-1.10.2
and I also got the same error message.
from cntk.
Related Issues (20)
- Value goes invalid when using TestMinibatch
- Request for a no-opencv dotnet release
- CNTK C# Crash when layer is deeper
- Convolution 1D CNTK C++ HOT 1
- Add support to release linux aarch64 wheels
- Error while deploying MS Teams Bot with SSO
- how to install in python3.8
- A model causes CNTK crash with cudnnSetPoolingNDDescriptor when invoking the pooling operator
- This repo is missing important files HOT 1
- API document issue due to syntax in source code HOT 1
- Error: could not find all specified 'to_nodes' in clone. Looking for ['relu5'], found [None]
- SequenceClassification.py does not work out of box HOT 1
- Microsoft has absolutely no potential
- development of a neural network for object search
- join vcpkg
- cntk crash and pycharm process finished with exit code -1066598274 (0xC06D007E)
- ConvolutionTranspose2D outputs normally when num_filters is tuple and dilation is 0
- program crash when get gradient of `ConvolutionTranspose2D`
- when shape contains negative integer, input_variable should throw an error HOT 1
- MAX_POOLING crash, when pooling_window_shape contains 0 HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from cntk.