I've taken some pains to write a bunch of boilerplate that will make working with models and data much easier. This has the unfortunate consequence of making the code look more complicated when you're new to looking at it. I hope I've struck a decent balance between simplifying the initial introduction and simplifying the workflow later on when you realize you need more/better tools.
See small_network.py
for a simple example. See wavelet_example.py
for an
intermediate example.
-
I'm using Python 3.6.10. Other versions not tested but probably fine.
-
PyTorch: On a Mac this simply looks like
pip install torch torchvision
(with nocuda
installation).- Note: at time of writing,
torch
does not cooperate with PIL version 7.0.0. To remedy this, you can runpip install 'PIL<7.0.0
- Note: at time of writing,
-
PyWavelets:
pip install PyWavelets
-
sklearn:
pip install -U scikit-learn
-
numpy
,pandas
,matplotlib
-
torchsummary
:pip install torchsummary
Fork the intro_pytorch repo. When you're in the forked repo on your own GitHub profile, click the green "Clone or download" button and then click the copy-to-clipboard icon. Next, open up a terminal window, navigate to your favourite directory and run:
git clone `pbpaste`
cd intro_pytorch/
jupyter notebook
If you're not using a Mac, then substitute pbpaste
for the link that was
copied to your clipboard.