TQT

TQT's pytorch implementation. Note, the Vitis Implement of TQT has different methods for the numbers.py to match with the DPU.

Notice

Now availabel at pip install tqt!

Networks quantized via this package could be find at https://github.com/PannenetsF/QuantizationPool.

Contributing

It will be great of you to make this project better! There is some ways to contribute!

1. To start with, issues and feature request could let maintainers know what's wrong or anything essential to be added.
2. If you use the package in you work/repo, just cite the repo and add a dependency note!
3. You can add some function in torch.nn like HardTanh and feel free to open a pull request! The code style is simple as here.

Acknowledgment

The initial version of tqt-torch is developed by Jinyu Bai.

The beta version was tested by Jinghan Xu, based on whose feedback a lot of bugs were fixed.

The original papar could be find at Arxiv, Trained Quantization Thresholds for Accurate and Efficient Fixed-Point Inference of Deep Neural Networks.