mdsuite.transformations.transformations module¶
MDSuite: A Zincwarecode package.
License¶
This program and the accompanying materials are made available under the terms of the Eclipse Public License v2.0 which accompanies this distribution, and is available at https://www.eclipse.org/legal/epl-v20.html
SPDX-License-Identifier: EPL-2.0
Copyright Contributors to the Zincwarecode Project.
Contact Information¶
email: zincwarecode@gmail.com github: https://github.com/zincware web: https://zincwarecode.com/
Citation¶
If you use this module please cite us with:
Summary¶
Parent class for the transformations.
- exception mdsuite.transformations.transformations.CannotFindPropertyError[source]¶
Bases:
Exception
Exception for when a property cannot be found.
- exception mdsuite.transformations.transformations.CannotFindTransformationError[source]¶
Bases:
Exception
Exception for when a transformation cannot be found.
- class mdsuite.transformations.transformations.MultiSpeciesTrafo(input_properties: Optional[Iterable[PropertyInfo]] = None, output_property: Optional[PropertyInfo] = None, scale_function=None, dtype=tf.float64)[source]¶
Bases:
Transformations
Base class for all transformations, where information of multiple species is combined in the transformation of a new property.
- run_transformation(species: Optional[Iterable[str]] = None) None [source]¶
Perform the batching and data loading for the transformation, then calls transform_batch :param species: Names of the species on which to perform the transformation. :type species: Iterable[str]
- abstract transform_batch(batch: Dict[str, Dict[str, tf.Tensor]], carryover: Any = None) tf.Tensor | Tuple[tf.Tensor, Any] [source]¶
Do the actual transformation.
- Parameters:
batch (dict) – The batch to be transformed. structure is {‘Species1’: {‘Property1’: tensordata, …}, …}
carryover (any) – if the transformation batching is only possible with carryover, this argument will provide it, see below.
- Returns:
Either the transformed batch (tf.Tensor)
Or tuple of (<transformed batch>, <carryover>),
where the carryover can have any type.
The carryover will be used as the optional argument for the next batch
- class mdsuite.transformations.transformations.SingleSpeciesTrafo(input_properties: Optional[Iterable[PropertyInfo]] = None, output_property: Optional[PropertyInfo] = None, scale_function=None, dtype=tf.float64)[source]¶
Bases:
Transformations
Base class for transformations where the transformation is applied to each species separately.
- run_transformation(species: Optional[Iterable[str]] = None)[source]¶
Perform the batching and data loading for the transformation, then calls transform_batch :param species: Names of the species on which to perform the transformation. :type species: Iterable[str]
- abstract transform_batch(batch: Dict[str, Tensor], carryover: Optional[Any] = None) Union[Tensor, Tuple[Tensor, Any]] [source]¶
Do the actual transformation.
- Parameters:
batch (dict) – The batch to be transformed. structure is {‘Property1’: tansordata, …}
carryover (any) – if the transformation batching is only possible with carryover, this argument will provide it.
- Returns:
Either the transformed batch (tf.Tensor)
Or tuple of (<transformed batch>, <carryover>),
where the carryover can have any type.
The carryover will be used as the optional argument for the next batch
- class mdsuite.transformations.transformations.Transformations(input_properties: Optional[Iterable[PropertyInfo]] = None, output_property: Optional[PropertyInfo] = None, scale_function=None, dtype=tf.float64)[source]¶
Bases:
object
Parent class for MDSuite transformations.
- data_manager¶
data manager for handling the data transfer
- Type:
- memory_manager¶
memory manager for the computation.
- Type:
- property database¶
Update the database.
replace for https://github.com/zincware/MDSuite/issues/404
- property experiment: Experiment¶
//github.com/zincware/MDSuite/issues/404.
- Type:
TODO replace for https