Agat

- Using self-defined tf-based functions to calculate Pearson r: [agat/lib/GatLib.pyL248-L259](https://github.com/jzhang-github/AGAT/tree/v7.13.3/agat/lib/GatLib.py#L248-L259)

This self-defined function can handle `ValueError: array must not contain infs or NaNs`.

- Fix a bug: [bug](https://github.com/jzhang-github/AGAT/tree/v7.13.3/agat/model/ModelFit.py#L280)
- Clip optimizer grads: [clipnorm=1.0](https://github.com/jzhang-github/AGAT/tree/v7.13.3/agat/default_parameters.py#L88-89)

- Fix bugs in high-throughput predict:
- [agat/app/cata/generate_adsorption_sites.py](https://github.com/jzhang-github/AGAT/tree/v7.13.2/agat/app/cata/generate_adsorption_sites.py#L218)
- [agat/app/cata/high_throughput_predict.py](https://github.com/jzhang-github/AGAT/tree/v7.13.2/agat/app/cata/high_throughput_predict.py#L207)
- [agat/app/cata/high_throughput_predict.py](https://github.com/jzhang-github/AGAT/tree/v7.13.2/agat/app/cata/high_throughput_predict.py#L250)
- [agat/app/cata/high_throughput_predict.py](https://github.com/jzhang-github/AGAT/tree/v7.13.2/agat/app/cata/high_throughput_predict.py#L291)
- [agat/app/GatApp.py](https://github.com/jzhang-github/AGAT/tree/v7.13.2ain/agat/app/GatApp.py#L69-L70)
- [agat/default_parameters.py](https://github.com/jzhang-github/AGAT/tree/v7.13.2/agat/default_parameters.py#L133)

- Deprecate redundant training configurations:
- `train_energy_model`: [agat/model/ModelFit.py](https://github.com/jzhang-github/AGAT/tree/v7.13.2/agat/model/ModelFit.py#L96) and [agat/model/ModelFit.py](https://github.com/jzhang-github/AGAT/tree/v7.13.2/agat/model/ModelFit.py#L198)
- `train_force_model`: [agat/model/ModelFit.py](https://github.com/jzhang-github/AGAT/tree/v7.13.2/agat/model/ModelFit.py#L274) and [agat/model/ModelFit.py](https://github.com/jzhang-github/AGAT/tree/v7.13.2/agat/model/ModelFit.py#L398)
- `new_energy_train`
- `new_force_train`
- `load_graphs_on_gpu`

- Fix a bug here: [agat/model/ModelFit.py](https://github.com/jzhang-github/AGAT/tree/v7.13.1/agat/model/ModelFit.py#L243)
- Load well-trained models: [agat/model/GatEnergyModel.py](https://github.com/jzhang-github/AGAT/tree/v7.13.1/agat/model/GatEnergyModel.py#L154-L197) and [agat/model/GatForceModel.py](https://github.com/jzhang-github/AGAT/tree/v7.13.1/agat/model/GatForceModel.py#L201-L246)
- Test with best model after training. [agat/model/ModelFit.py](https://github.com/jzhang-github/AGAT/tree/v7.13.1/agat/model/ModelFit.py#L222) and [agat/model/ModelFit.py](https://github.com/jzhang-github/AGAT/tree/v7.13.1/agat/model/ModelFit.py#L422).

- Raise exception if error occurs when parsing OUTCAR file. [agat/data/data.py](https://github.com/jzhang-github/AGAT/tree/v7.13/agat/data/data.py#L595-L599)
- Remove `os` from the root name space. [agat/__init__.py](https://github.com/jzhang-github/AGAT/tree/v7.13/agat/__init__.py#L18)
- Fix a bug when build graphs. See [agat/data/data.py](https://github.com/jzhang-github/AGAT/tree/v7.13/agat/data/data.py#L400-L411) and [agat/data/data.py](https://github.com/jzhang-github/AGAT/tree/v7.13/agat/data/data.py#L444-L452). Specifically, one needs to cast `tf.tensor` as `np.array` before building graph properties with a very large tensor. [agat/data/data.py](https://github.com/jzhang-github/AGAT/tree/v7.13/agat/data/data.py#L408-L409).
- Debug at these lines of [agat/data/data.py](https://github.com/jzhang-github/AGAT/tree/v7.13/agat/data/data.py): [L553](https://github.com/jzhang-github/AGAT/tree/v7.13/agat/data/data.py#L553) and [L585-L588](https://github.com/jzhang-github/AGAT/tree/v7.13/agat/data/data.py#L585-L588).

- Using relative import. For example: [agat/__init__.py](https://github.com/jzhang-github/AGAT/tree/v7.12.2/agat/__init__.py#L14-L16)
- Update [documentations](https://jzhang-github.github.io/AGAT/).
- Import useful objects only. For example: [agat/app__init__.py](https://github.com/jzhang-github/AGAT/tree/v7.12.2/agat/app/__init__.py#L11)
- Return test MAE after training. [agat/model/ModelFit.py](https://github.com/jzhang-github/AGAT/tree/v7.12.2/agat/model/ModelFit.py#L247) and [agat/model/ModelFit.py](https://github.com/jzhang-github/AGAT/tree/v7.12.2/agat/model/ModelFit.py#L442)

- Import `pymatgen` module when necessary. See [agat/data/AtomicFeatures.py](https://github.com/jzhang-github/AGAT/tree/v7.12.1/agat/data/AtomicFeatures.py#L11). This feature was changed back.
- Specify device when building graphs. See [agat/app/GatApp.py](https://github.com/jzhang-github/AGAT/tree/v7.12.1/agat/app/GatApp.py#L69), [agat/data/data.py](https://github.com/jzhang-github/AGAT/tree/v7.12.1/agat/data/data.py#L79-L83)
- Add default gpu specification when building database. [agat/default_parameters.py](https://github.com/jzhang-github/AGAT/tree/v7.12.1/agat/default_parameters.py#L60)
- Attache distributions at [dist](https://github.com/jzhang-github/AGAT/tree/v7.12.1/dist).