Jupyter notebook for deep learning workshop at MoProc 2019. To open in binder, click on the binder badge next to the version you want. The "MoProc 2019" version will be the same as the notebook presented at the workshop, and the "latest" version will have any upates since the workshop.

Latest version

The latest version includes a small edit about global convergence, more information on using CuArrays for GPU acceleration, and a batching bugfix.

MoProc 2019 version

Instructions for running the notebook locally

These instructions will assume you have already installed Julia. They will also default you to using the latest version. If you would like to use the conference version, you will have to change branches on this page (in the upper left corner) before downloading, or change branches using Git if you use Git to clone the repository.

1. Download this repository. If you use Git, this can be done using git clone https://github.com/maetshju/ling_513_dnn_lecture. Otherwise, use the green "Clone or download" button at the upper right of the page.

2. Navigate into the cloned repository, and then run Julia in the current environment:

   julia --project=.

3. Install the appropriate packages in Julia:

] instantiate

The "]" at the beginning means to press the "]" key to enter the package interface. Copying the whole line of code including the "]" in the Julia REPL console should automatically enter the package interface as well.

4. Download and unzip the subset of the MALD data set that we will be working with. If you have access to a Bash console, you can run the "postBuild" script to automatically download and extract the files.

sh postBuild

If not, download them from the University of Alberta institutional repository. Make sure that the folder containing the MALD folders is placed in the same directory as the rest of the files in this repository.

5. Start a notebook in Julia. At the console interface:

using IJulia
notebook()

Some information will be printed to the console, and a Jupyter notebook window will appear in your default web browser. If it asks you for a token, on a regular console (or CMD window on Windows), you will need to run jupyter notebook list, which will print out a URL with the token in it, which can be pasted into the webpage.

5. Use the Jupyter interface to find the corresponding "ipynb" files you downloaded as part of this repository. Clicking on them will open the notebooks, which you can then run from there.