2020/08/14
- Add model dynamic quantization.
2020/08/12
- Add model profiling.
2020/08/02
- Use batch first and use pad_sequence.
- Enable model checkpoint save and load.
2020/07/31
- Enable bi-direction and multi-layer RNN.
2020/07/28
- Deprecate TextCNNConfig and RNNConfig by making Config flexible to accept arbitrary params.
2020/07/26
- Allow config to set arbitrary values
2020/07/25
- Make RNN classification model configurable to use RNN, LSTM, and GRU
- Make RNN classification model configurable on the number of layers
2020/07/24
- Refactored the file structure by separating configs, models, and training logic
2020/07/23
- Fix the bug that the weights conv layers are not correctly updated during back-propagation
2020/07/22
- Make TextCNN graph configurable on number of conv branches