Installing DBNsim

Requirements

For running DBNsim on you own server you will need:

Please note that you have to use Python 2 (preferably 2.7), not Python 3! This is beacause, unfortunately, DBNsim currently depends on two libraries that are only available for Python 2.

If you have installed Python under Windows but your command line says it doesn't recognize the commands pip and python, then you just have to add Python to your path. This is generally done by going to System Properties / Advanced / Environment Variables.

Basic installation

For installing DBNsim (both for CPU and GPU usage) follow these steps:

git clone https://github.com/ggiuffre/DBNsim.git
cd DBNsim
pip install --user -r requirements.txt

In addition, it is recommended to install scipy, with the command pip install --user scipy. Windows user may find it difficult but this is currently an optional dependency; the only thing is that without it you won't be able to handle MATLAB files as an administrator (see here).

If you don't have Git installed, you can still go to the GitHub repo, click "Clone or Download", then "Download ZIP"; now extract the archive, and then cd into the extracted directory.

Adding GPU support

If you have a CUDA-enabled GPU, you can install CUDAMat. To install it, first exit the DBNsim directory (this is very important) — a simple way to exit the directory is to cd into your home directory; then execute the following commands:

git clone https://github.com/cudamat/cudamat.git
cd cudamat
pip install --user .

As before, if you don't have Git installed you can download a zipped version of CUDAMat from GitHub.

If you get errors while installing with pip, make sure you have the CUDA drivers installed, along with the NVIDIA CUDA compiler (nvcc). If you're using Ubuntu, then you can install them with:

sudo apt-get install nvidia-cuda-toolkit

After you manage to install CUDAMat, remember to test it. To do this, exit the cudamat directory (this is very important), enter a Python interpreter and type:

import numpy as np
import cudamat as cm

cm.cublas_init()

a = cm.CUDAMatrix(np.random.rand(32, 256))
b = cm.CUDAMatrix(np.random.rand(256, 32))

c = cm.dot(a, b)
d = c.sum(axis = 0)

print(d.asarray())

Launching DBNsim

Finally, launch DBNsim with:

cd path/to/DBNsim
python DBNsite/manage.py runserver

... this actually launches DBNsim only locally. Instead, if you want other users to access it, type:

cd path/to/DBNsim
python DBNsite/manage.py runserver your.public.ip.address:8000

You should see something like:

Performing system checks...

loading data from pickle...
loading data from pickle...
loading data from pickle...
loading data from pickle...
System check identified no issues (0 silenced).
June 22, 2017 - 16:02:15
Django version 1.11.2, using settings 'DBNsite.settings'
Starting development server at http://your.public.ip.address:8000/
Quit the server with CONTROL-C.

This means the server is up and running. Now you should find the user interface at http://127.0.0.1:8000 if you're running the app locally. Or, if you're publishing it to a network, you should find it at http://your.public.ip.address:8000. 8000 is just an example — you can run DBNsim on any port you have access to.

Once you're on the page, you will be asked for a password. If the server administrator didn't change the default configuration, the password is user.

When you want to shut down the server, return to the command line interface and press CONTROL-C.

Changing the default configuration

As the server administrator, you have more power than normal users. It is possible to configure DBNsim under the following aspects:

For more about how to change the configuration, please read here.