{"id":3111,"date":"2016-06-15T17:57:29","date_gmt":"2016-06-15T17:57:29","guid":{"rendered":"http:\/\/ri.itservices.manchester.ac.uk\/csf-apps\/?page_id=3111"},"modified":"2017-03-30T10:49:00","modified_gmt":"2017-03-30T10:49:00","slug":"theano","status":"publish","type":"page","link":"https:\/\/ri.itservices.manchester.ac.uk\/csf-apps\/software\/applications\/theano\/","title":{"rendered":"Theano"},"content":{"rendered":"

Overview<\/h2>\n

Theano<\/a> is a Python library that allows you to define, optimize, and evaluate mathematical expressions involving multi-dimensional arrays efficiently.<\/p>\n

Version 0.8.2 is installed on the CSF. It uses the CUDA 5.5.22 library on the GPU nodes.<\/p>\n

Restrictions on use<\/h2>\n

There are no restrictions on running Theano. However, access to the GPU nodes is not possible until you have been added to the correct group. Please contact its-ri-team@manchester.ac.uk<\/a> to request this.<\/p>\n

You are encouraged to cite the library in any research. Please see the citation page<\/a> for details.<\/p>\n

Set up procedure<\/h2>\n

To access the software you must first load the modulefile:<\/p>\n

\r\nmodule load apps\/binapps\/theano\/0.8.2\r\n<\/pre>\n
This will automatically load the Anaconda v2.5.0 Python module files (which provides python 2.7.11) and also the CUDA 5.5.22 modulefile.<\/p>\n
If you require the PyCUDA libraries please also load the following additional modulefile:<\/p>\n
\r\nmodule load libs\/gcc\/pycuda\/2016.1.1\r\n<\/pre>\n
Running the application<\/h2>\n
Please do not run Theano on the login node. Jobs should be submitted to the compute nodes via batch.<\/p>\n
Serial batch job submission – GPU<\/h3>\n
Make sure you have the modulefile loaded then create a batch submission script, for example:<\/p>\n
\r\n#!\/bin\/bash\r\n#$ -S \/bin\/bash\r\n#$ -cwd             # Job will run from the current directory\r\n#$ -V               # Job will inherit current environment settings\r\n##### Use ONE of the following to select a GPU node\r\n#$ -l nvidia       # Nvidia M2050\r\n#$ -l nvidia_k20   # Nvidia K20 (restricted to specific research group)\r\n\r\npython myscript.py\r\n<\/pre>\n
Submit the jobscript using: <\/p>\n
qsub scriptname<\/em><\/pre>\n
where scriptname<\/em> is the name of your jobscript.<\/p>\n
Parallel batch job submission – CPU<\/h3>\n
Make sure you have the modulefile loaded then create a batch submission script, for example:<\/p>\n
\r\n#!\/bin\/bash\r\n#$ -S \/bin\/bash\r\n#$ -cwd             # Job will run from the current directory\r\n#$ -V               # Job will inherit current environment settings\r\n#$ -pe smp.pe 4     # 4 core job\r\n\r\n# The next line ensures that the software automatically uses the no. of cores requested (in this case 4)\r\n\r\nexport OMP_NUM_THREADS=$NSLOTS\r\n\r\npython myscript.py\r\n<\/pre>\n
Interactive Usage<\/h3>\n
To use python interactively first start an interactive session<\/em> on a GPU node using:<\/p>\n
\r\n# Access to M2050 GPU\r\nqrsh -l inter -l nvidia\r\n\r\n# Access to K20 GPU\r\nqrsh -l inter -l nvidia_k20\r\n\r\n# When logged in to a GPU node, set up as before:\r\nmodule load apps\/binapps\/theano\/0.8.2\r\npython\r\n>>> import theano\r\n>>> ... python code...\r\n\r\n\/\/ When finished:\r\nCtrl-d\r\n\r\n# Now end your interactive session to return to the CSF login node\r\nexit\r\n<\/pre>\n
Further info<\/h2>\n