27e9453d19
* add code comments * merge *_saver.py & *_loader.py in io/ * (ancient codes) rename Loss into LossFromTorch |
||
|---|---|---|
| .. | ||
| config.cfg | ||
| dataloader.py | ||
| example.py | ||
| main.py | ||
| predict.py | ||
| prepare.py | ||
| README.md | ||
| Word2Idx.py | ||
Prototype
Word2Idx.py
A mapping model between words and indexes
embedding.py
embedding modules
Contains a simple encapsulation for torch.nn.Embedding
encoder.py
encoder modules
Contains a simple encapsulation for torch.nn.LSTM
aggregation.py
aggregation modules
Contains a self-attention model, according to paper "A Structured Self-attentive Sentence Embedding", https://arxiv.org/abs/1703.03130
predict.py
predict modules
Contains a two layers perceptron for classification
example.py
An example showing how to use above modules to build a model
Contains a model for sentiment analysis on Yelp dataset, and its training and testing procedures. See https://arxiv.org/abs/1703.03130 for more details.
prepare.py
A case of using Word2Idx to build Yelp datasets
dataloader.py
A dataloader for Yelp dataset
It is an iterable object, returning a zero-padded batch every iteration.