This is the beta version of DeepOBS with **PyTorch** and TensorFlow support as well as an added tuning module.
Major Changes
- A PyTorch implementation (though not for all test problems yet). - A refactored Analyzer module (more flexibility and interpretability). - A Tuning module that automates the tuning process. - A separate validation set for hyperparameter tuning.
And many more changes which can be gathered from the <a href="https://github.com/fsschneider/DeepOBS/blob/v1.2.0-beta0/CHANGELOG.md">Changelog</a></li>.
Many thanks to [Aaron Bahde](https://github.com/abahde/) for spearheading the development of DeepOBS 1.2.0.
1.1.2
This mostly includes a small to fix the installed version of matplotlib2tikz is 0.6.18.
All results are still comparable from version 1.1.1 and 1.1.0.
1.1.1
The baselines are now hosted in a separate repository, which allows our package to be pip installable using
pip install deepobs.
Nothing else has changed, all results are still comparable from version 1.1.0
1.1.0
We have re-written most parts of the code to make it more flexible and easier to use.