The repo of the of the benchmark is:
http://github.com/jaberg/DeepLearningBenchmarks

The current script to run the tests is in this file theano/run.sh. Some tests are commented. We should run all cases in the buildbot: mlp, convnet, rbm on cpu and gpu.

Can you do as in the DeepLearningTutorial in this repo:
git@github.com:lisa-lab/DeepLearningTutorials.git

The script that do the speed execution is in the file code/test.py in the function speed. We should print the speed information as in this function as this is processed by another script.