python cntkDemo.py error on Ubuntu 14.04

[codelast]

When running command python cntkDemo.py on Ubuntu 14.04, I got following errors:

OpenBLAS : Your OS does not support AVX instructions. OpenBLAS is using Nehalem kernels as a fallback, which may give poorer performance.
Traceback (most recent call last):
File "/home/codelast/.miniconda3/envs/py36/lib/python3.6/site-packages/cntk/cntk_py.py", line 18, in swig_import_helper
return importlib.import_module(mname)
File "/home/codelast/.miniconda3/envs/py36/lib/python3.6/importlib/init.py", line 126, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
File "", line 978, in _gcd_import
File "", line 961, in _find_and_load
File "", line 948, in _find_and_load_unlocked
ModuleNotFoundError: No module named 'cntk._cntk_py'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "cntkDemo.py", line 7, in
import cntk_to_ell
File "./../../../../build/tools/importers/CNTK/cntk_to_ell.py", line 17, in
from cntk.layers import Convolution, MaxPooling, AveragePooling, Dropout, BatchNormalization, Dense
File "/home/codelast/.miniconda3/envs/py36/lib/python3.6/site-packages/cntk/init.py", line 10, in
from . import cntk_py
File "/home/codelast/.miniconda3/envs/py36/lib/python3.6/site-packages/cntk/cntk_py.py", line 21, in
_cntk_py = swig_import_helper()
File "/home/codelast/.miniconda3/envs/py36/lib/python3.6/site-packages/cntk/cntk_py.py", line 20, in swig_import_helper
return importlib.import_module('_cntk_py')
File "/home/codelast/.miniconda3/envs/py36/lib/python3.6/importlib/init.py", line 126, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
ImportError: libmpi_cxx.so.1: cannot open shared object file: No such file or directory

And before that, I've successfully executed the following steps as the doc asked:

(1) make ELL
(2) generate Python bindings & run the test, below are my output:

OpenBLAS : Your OS does not support AVX instructions. OpenBLAS is using Nehalem kernels as a fallback, which may give poorer performance.
Testing HingeLoss.Evaluate(2, 1) ... Passed
Testing HingeLoss.Evaluate(-2, -1) ... Passed
Testing HingeLoss.Evaluate(2, -1) ... Passed
Testing HingeLoss.Evaluate(-2, 1) ... Passed
Testing HingeLoss.Evaluate(0, 1) ... Passed
Testing HingeLoss.Evaluate(0, -1) ... Passed
Testing HingeLoss.GetDerivative(2, 1) ... Passed
Testing HingeLoss.GetDerivative(-2, -1) ... Passed
Testing HingeLoss.GetDerivative(2, -1) ... Passed
Testing HingeLoss.GetDerivative(-2, 1) ... Passed
Testing HingeLoss.GetDerivative(0, 1) ... Passed
Testing HingeLoss.GetDerivative(0, -1) ... Passed
Testing LogLoss.Evaluate(2, 1) ... Passed
Testing LogLoss.Evaluate(-2, -1) ... Passed
Testing LogLoss.Evaluate(2, -1) ... Passed
Testing LogLoss.Evaluate(-2, 1) ... Passed
Testing LogLoss.Evaluate(0, 1) ... Passed
Testing LogLoss.Evaluate(0, -1) ... Passed
Testing LogLoss.GetDerivative(2, 1) ... Passed
Testing LogLoss.GetDerivative(-2, -1) ... Passed
Testing LogLoss.GetDerivative(2, -1) ... Passed
Testing LogLoss.GetDerivative(-2, 1) ... Passed
Testing LogLoss.GetDerivative(0, 1) ... Passed
Testing LogLoss.GetDerivative(0, -1) ... Passed
Testing SquaredLoss.Evaluate(4, 4) ... Passed
Testing SquaredLoss.Evaluate(4, 2) ... Passed
Testing SquaredLoss.Evaluate(2, 4) ... Passed
Testing SquaredLoss.GetDerivative(4, 4) ... Passed
Testing SquaredLoss.GetDerivative(4, 2) ... Passed
Testing SquaredLoss.GetDerivative(2, 4) ... Passed
functions_test passed
[1] One subgraph
Subgraph Vertices Edges Cycles
0 6 8 3
[2] One subgraph
Subgraph Vertices Edges Cycles
0 6 6 1
[3] One subgraph
Subgraph Vertices Edges Cycles
0 8 11 4
[tree_0] One subgraph
Subgraph Vertices Edges Cycles
0 17 18 2
[tree_1] One subgraph
Subgraph Vertices Edges Cycles
0 45 54 10
[tree_2] One subgraph
Subgraph Vertices Edges Cycles
0 73 90 18
[tree_3] One subgraph
Subgraph Vertices Edges Cycles
0 101 126 26
model_test passed
Model 1 size: 6
Model 2 size: 6
Model 3 size: 8
Tree 0 size: 17
Tree 1 size: 45
Tree 2 size: 73
Tree 3 size: 101
Loading file ../../../examples/data/model_1.model
Loading file ../../../examples/data/model_2.model
common_test passed
trainers_test.test -- TBD
trainers_test passed
predictors_test.test -- TBD
predictors_test passed
nodes_test.test -- TBD
nodes_test passed
linear_test.test -- TBD
linear_test passed
evaluators_test.test -- TBD
evaluators_test passed
Testing ModelBuilder ... Passed
modelbuilder_test passed

(3) install CNTK python packages:

(py36) [codelast@ gettingStarted]$ pip install https://cntk.ai/PythonWheel/CPU-Only/cntk-2.0-cp36-cp36m-linux_x86_64.whl
Requirement already satisfied: cntk==2.0 from https://cntk.ai/PythonWheel/CPU-Only/cntk-2.0-cp36-cp36m-linux_x86_64.whl in /home/codelast/.miniconda3/envs/py36/lib/python3.6/site-packages
Requirement already satisfied: numpy>=1.11 in /home/codelast/.miniconda3/envs/py36/lib/python3.6/site-packages (from cntk==2.0)
Requirement already satisfied: scipy>=0.17 in /home/codelast/.miniconda3/envs/py36/lib/python3.6/site-packages (from cntk==2.0)

(4) download the CNTK ImageNet reference model

curl -O https://www.cntk.ai/Models/Caffe_Converted/VGG16_ImageNet_Caffe.model

So, could anyone give me a hint on how to solve the problem? Thanks a lot.

[lovettchris]

> ImportError: libmpi_cxx.so.1: cannot open shared object file: No such file or directory

I wonder if "sudo apt-get install libopenmpi-dev" will solve this problem?

There's a lot more CNTK Linux instructions if that doesn't work, one would assume the conda cntk python package should have taken care of all these prerequisites, but perhaps it has a bug...

[codelast]

@lovettchris Thank you for providing the info, I'll try it later.

[codelast]

@lovettchris
sudo apt-get install libopenmpi-dev can't solve the problem because on Ubuntu 14.04 the package installed by this way is too old:

[codelast@ ~]$ apt-cache show libopenmpi-dev
Version: 1.6.5-8

[codelast@ ~]$ ll /usr/lib/libmpi.so*
lrwxrwxrwx 1 root root 27 7 12 00:03 /usr/lib/libmpi.so -> /etc/alternatives/libmpi.so
lrwxrwxrwx 1 root root 15 12 29 2013 /usr/lib/libmpi.so.1 -> libmpi.so.1.0.8
lrwxrwxrwx 1 root root 27 12 29 2013 /usr/lib/libmpi.so.1.0.8 -> openmpi/lib/libmpi.so.1.0.8

While ELL requires libmpi.so.12, so I have to do the following things:

(1) uninstall all the openmpi-related packages on my Ubuntu 14.04:

sudo apt-get remove libmpich-dev mpi-default-dev libopenmpi-dev libopenmpi1.6 openmpi-common

(2) download the openmpi-1.10.2.tar.gz package on this web page, and then :

tar zxf openmpi-1.10.2.tar.gz
cd openmpi-1.10.2/
./configure
make
sudo make install

(3) check the installation:

[codelast@ ~]$ sudo find ./ -name libmpi.so.12
./usr/local/lib/libmpi.so.12

That's it! Problem solved!