Run the oximachine
Project description
oximachinerunner
oximachine for AiiDA lab: Core functionalities of oximachine with stripped dependencies.
Warning: This model works excellent on a test set but it might give fully unphysical predictions in some cases. Consider it in alpha phase
-
It is good to know where it fails
-
We work on improving the model by training it on a larger subset of the CSD with a new architecture
-
The featurization can be slow in some cases. In practice, it is best to get the smallest possible cell of a clean structure.
-
There is still one dependency on one of my forks of a well-known package.
-
The package is slow
-
For compatability and reproducibility we need to pin an old scikit-learn version
features not deployed for AiIDA lab
- Feature importance (slow as it has to integrate the dataset. Also, it is quite likely that we will break the API here in the future when we add new features)
- Most similar structures in training set (is typically fast though, as it uses a KDtree)
- Uncertainity estimate (not sure how the best way to use this in a workchain?)
assets
votingclassifier.joblib
is the model that is currently deployed. It is a voting classifier with four different base estimatorsscaler.joblib
is the standard scalerKAJZIH_freeONLY.cif
andACODAA.cif
are some test structures.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for oximachinerunner-0.2a0-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 3ac023ac4a689c91e5076faca5089493a14a7d17a04557b725c5468da8d9d4e6 |
|
MD5 | 606e0bdf2f16d43283662e699fc5cb96 |
|
BLAKE2b-256 | b6216f5fa8d3c19db8558c4a6a0400a9fc91f5b320110271c20cf73bf8927e98 |