← NLP · Repository
This track compares recurrent sequence classifiers sharing a local configuration and training shape: embedding input, recurrent encoder, classifier head, and cross-entropy optimisation.
| Folder | Architecture | TensorFlow construction | Dataset status |
|---|---|---|---|
1_VanillaRNN |
Vanilla RNN | build_vanilla_rnn |
Adapter base is incomplete |
2_LSTM |
LSTM | build_lstm |
Adapter base is incomplete |
3_GRU |
GRU | build_gru |
Adapter base is incomplete |
4_BidirectionalRNN |
Bidirectional RNN | build_bidirectional_rnn |
Adapter base is incomplete |
5_StackedRNN |
Stacked RNN | build_stacked_rnn |
Adapter base is incomplete |
The source references AG News, DBPedia, and IMDB adapters and uses tf.data
in dataset code. However, each dataset.py contains an explicit
NotImplementedError; architecture pages therefore describe the model but do
not claim an end-to-end verified dataset run.