System version: Red Hat 4.4.6-4
A. If not networked, select Local installation.
Install Theano's dependent libraries first, including:
scipy-0.16.1
numpy-1.9.2
nose-1.3.7 (optional)
lapack-3.5.0
atlas-3.10.2
sklearn-0.16.1 (optional)
1. Installation of Blas, Blas-devel, Lapack, Lapack-devel, Atlas, Atlas-devel
Can be directly: Yum install Blas, Blas-devel, Lapack, Lapack-devel, Atlas, Atlas-devel
PS: If you do not have root privileges, you need to copy the source code in sequence, execute Python setup.py install
2. Install NumPy scipy
3. Installing Theano
Download Theano:https://github.com/theano/theano
Unzip && Pyhton setup.py install && Pyhton setup.py Develop
To modify the environment variables, you can find the Theano installation address in Python:
VI ~/.BASHRC
Added: Export pythonpath=/usr/local/lib/python2.7/site-packages: $PYTHONPATH
SOURCE ~/.BASHRC
Two. Machine Networking
First step: Yum install Blas, Blas-devel, Lapack, Lapack-devel, Atlas, Atlas-devel
Step two: Pip install NumPy, pip install scipy
Step three:pip install Theano
Three. Problems encountered
When import Theano, some errors are sometimes returned
Problem 1:fpic problem
Workaround: Reinstall Python and compile as follows:
./configure--enable-shared--prefix=/usr/localmake doinstall
Problem 2:gof problem
Workaround: Install multiple Theano, view files under/usr/local/lib/python2.7/site-packages/, find multiple versions of Theano, and keep one.
PS: No root privileges
Four. Articles that are considered helpful by individuals
http://blog.csdn.net/a6210575/article/details/46991937
Http://www.cnblogs.com/lanxuezaipiao/archive/2012/10/21/2732864.html
Http://www.360doc.com/content/12/0322/18/6938655_196712249.shtml
http://fhqdddddd.blog.163.com/blog/static/18699154201432844414757/
Theano installation and GPU acceleration under CentOS 6