This is a collection of PyTorch examples which I'm developing while learning and transitioning to PyTorch. The examples demonstrate different model architectures and tasks. Each example is commented rather heavily to make them useful to other fellow learners. Each example is also self-contained by design. There is no shared code. The datasets are not synthetic. I try to use popular, real-world datasets that hopefully make the examples more relevant. Lastly, the models are not optimized for performance but focus primarily on demonstration and teaching.
| File | Model | Task | Dataset |
|---|---|---|---|
mlp_imgcls_mnist.py |
MLP | Image classification | Mnist (torchvision) |
mlp_tblbcls_space.py |
MLP | Table binary classification | Spaceship Titanic (Kaggle) |
rnn0_lm_wikitext.py |
RNN (from scratch) | Language modeling | wikitext (txt files) |
rnn_lm_wikitext.py |
RNN | Languge modeling | wikitext (txt files) |
lstm_lm_wikitext.py |
LSTM | Language modeling | wikitext (txt files) |
lstm_ts_ |
LSTM | Timeseries prediction | |
lstm_ts_ |
LSTM | Timeseries prediction (multivariate) |
My intent is that these examples are useful to others, and that they demonstrate broad range of concepts, architectures, and tasks. It's of course tricky to decide what and how much to comment. The important goal to keep in mind is that the examples here shouldn't be the only resource while learning PyTorch so they don't have to explain everything.
If you think some comments or code is unclear, can be improved, or added please feel free to submit an issue.