Dear all,
the first lab will be held on Tuesday 16/03.
It'll be a hands-on introduction to PyTorch tensors and neural network modules.
I prepared this lab expecting all the students to have some knowledge of basic Python and NumPy (by which this lab will be heavily inspired). Also, some basic knowledge of Git is recommended but not required. Please contact me if you think you don't have the knowledge required.
The notebooks for the labs will be stored and updated on this GitHub repo.
If you wish to work locally (on your laptop or PC), please have PyTorch already installed for the first lab.
- to install Python we suggest installing it along with Anaconda (especially on Windows) as it comes with a number of pre-installed packages (not PyTorch unfortunately);
- to install PyTorch, refer to this page. If you have a CUDA-capable GPU and wish to use hardware acceleration during the lectures (NB: it isn't necessary), you'll need to pre-install CUDA.
- we suggest you install all the packages in a virtual environment (e.g., virtualenv or conda)
- moreover, you'll need some more packages which may not be installed on your PC or environment. You'll find a list (updated from lab to lab) in the file `requirements.txt` within the GitHub repo.
If you aren't sure about the steps above or do not wish to work locally, we warmly suggest using Google Colab: essentially it's a virtual machine (with the possibility to use GPUs as well) that has already NumPy and PyTorch pre-installed and it will save you a lot of time and head-scratches (and disk space as well). You'll need a Google account in order to use it.
Anyway, in case of doubts, just contact me (possibly before Tuesday) and I'll try to help you out.
Cheers!