Interpretdl

We release the version 0.5.1 of InterpretDL, with small fixes:

- Update readme, add the schema of relations among interpretation, interpretability and trustworthiness.
- Fix some imports errors.
- Add one more base Interpreter `IntermediateLayerInterpreter`.

Thanks Wgm-Inspur for correcting the parameter of `GradShapNLPInterpreter` used in tutorials.

We would also like to mention that the arguments `use_cuda` is deprecated. Use `device` directly.

We release the version 0.5.0 of InterpretDL, with new features as following:

- Two more evaluation metrics are added: Perturbation tests and PointGame, for measuring the trustworthiness of interpretation algorithms. APIs for [Perturbation](https://github.com/PaddlePaddle/InterpretDL/blob/master/interpretdl/evaluate_interpreter/perturbation.py), [PointGame](https://github.com/PaddlePaddle/InterpretDL/blob/master/interpretdl/evaluate_interpreter/localization.py#L24) and [PointGameSegmentation](https://github.com/PaddlePaddle/InterpretDL/blob/master/interpretdl/evaluate_interpreter/localization.py#L71) are available with corresponding tutorials, [1](https://github.com/PaddlePaddle/InterpretDL/blob/master/tutorials/perturbation_tutorial.ipynb) and [2](https://github.com/PaddlePaddle/InterpretDL/blob/master/tutorials/pointgame_tutorial.ipynb).
- Gradients are supported in the eval mode since Paddle2.2.1, which is supported too by InterpretDL, making the gradient computation easier.
- Deprecation of `use_cuda` is on the way. Use `device` directly.

We release the version 0.4.0 of InterpretDL, with new features as following:

- Add Consensus of Cross-Model Explanation Algorithm. See the [API](https://github.com/PaddlePaddle/InterpretDL/blob/master/interpretdl/interpreter/consensus.py) and the [tutorial](https://github.com/PaddlePaddle/InterpretDL/blob/master/tutorials/consensus_tutorial_cv.ipynb) for details.
- Add Deletion and Insertion Evaluation Algorithms, for measuring the trustworthiness of interpretation algorithms. See the [API](https://github.com/PaddlePaddle/InterpretDL/blob/master/interpretdl/evaluate_interpreter/deletion_insertion.py) and the [tutorial](https://github.com/PaddlePaddle/InterpretDL/blob/master/tutorials/del_ins_tutorial.ipynb) for details.
- Support Continuous Integration for code quality. We choose circleci for InterpretDL. The code coverage is 93% at this version.
- We add colorful badges in README ;)

We released the new version 0.3.0 of InterpretDL. The important changes are introduced as follows:

- Expose the API of image reader, where users can change the resize scale and the crop size.
- SmoothGrad optimized (V2).
- Documentation & Readme & Tutorials updates.

Small fixes and documentation updates.

We released the new version 0.2.4 of InterpretDL. The important changes are introduced as follows:
- Add three sub-abstract classes, [`InputGradientInterpreter`](https://github.com/PaddlePaddle/InterpretDL/blob/master/interpretdl/interpreter/abc_interpreter.py#L108), [`InputOutputInterpreter`](https://github.com/PaddlePaddle/InterpretDL/blob/master/interpretdl/interpreter/abc_interpreter.py#L206) and [`IntermediateLayerInterpreter `](https://github.com/PaddlePaddle/InterpretDL/blob/master/interpretdl/interpreter/abc_interpreter.py#L274). In each of them, we provide the PaddlePaddle implementation for getting `input gradients`, `output logits/probability` and `intermediate feature maps` respectively. In this way, the explanation algorithms can focus on the algorithmic implementations, leading to a good separation of framework and algorithms. Further developments would be easier. More sub-abstract classes including obtaining the gradients of intermediate layers will be implemented in the future.
- Add [TAMInterpreter](https://github.com/PaddlePaddle/InterpretDL/blob/master/interpretdl/interpreter/transition_attention_maps.py).
- In the instantiation of Interpreters, `use_cuda` would be deprecated soon. Use ``device`` directly.
- `_paddle_prepare` would be renamed to `_build_predict_fn`.
- Add unit tests.
- Improve documentation and small fixes.