Lis extensions for PyTorch
This project aims to provide useful extensions for PyTorch. As toolbox,
- It is NOT a API/model wrapper: Unlike Tensorflow, PyTorch's APIs are intuitive and easy to use. Hence I am going to make things complicated.
- It is written in pure python, which means you can use it over all platforms.
- It focuses on speed. All features are implemented with minimum operation and all calculations are vectorized.
- General Metrics:
- Accuracy (topk included)
- Error (topk included)
- Negative Log Liklyhood
- Mean Squared Error
- Mean IoU (intersection of union)
- Confusion Matrix
- General Meters
- AverageMeter
- SlidingWindowAverageMeter
- MovingAverageMeter
- SpeedMeter (under construction now)
- Onehot operator
- NLL over onehot
- Accuracy over onehot
- Embedding over onhot.
- Common Flags (under construction now)
Any issues, improvements or suggestsions are very welcome.