PyTorch Demo

This repository is built for the presentation given at the Future Talk 2019 event.

Prerequisites

For ease of use, I recommend you install the following software packages in advance.

Anaconda 3

Download and install the 3.x version of Anaconda. It comes with Python 3. After installation, make sure that the command 'conda' is a recognized command. If not check that INSTALL_DIR/anaconda3/bin is included in your path variable.

Create a new environment:

conda create -n pytorch-demo python=3.7

Activate the environment:

conda activate pytorch-demo

PyTorch

If you have a system with an NVIDIA graphics card that supports CUDA, choose the installation for GPU, otherwise install the CPU-only version.

GPU: conda install pytorch torchvision cudatoolkit=10.0 -c pytorch
CPU: conda install pytorch-cpu torchvision-cpu -c pytorch

This repository was prepared with PyTorch 1.1.0.

Jupyter Notebook

We will use Jupyter Notebook as an interactive code editor. First install it, then start:

conda install jupyter
jupyter notebook

This should open a new browser window.

Everything Else

You will need a bunch of other packages, nothing fancy. Just run

pip install -r requirements.txt

You are ready to start coding! Finally, when you are done working with your notebooks, deactivate the environment:

conda deactivate

Slides

The slides are available as an interactive Jupyter Notebook. If you want to use and adapt them yourself, you will need RISE:

conda install -c conda-forge rise

This is optional and not required to run and interact with the code.