gt4sd.algorithms.generation.torchdrug.core module¶

Torchdrug generation algorithm.

Summary¶

Classes:

`TorchDrugGCPN`	Interface for TorchDrug Graph-convolutional policy network (GCPN) algorithm.
`TorchDrugGenerator`
`TorchDrugGraphAF`	Interface for TorchDrug flow-based autoregressive graph algorithm (GraphAF).

Reference¶

class TorchDrugGenerator(configuration, target=None)[source]¶

Bases: GeneratorAlgorithm[S, None]

__init__(configuration, target=None)[source]¶

TorchDrug generation algorithm.

Parameters

configuration (AlgorithmConfiguration) – domain and application specification, defining types and validations. Currently supported algorithm versions are: “zinc250k_v0”, “qed_v0” and “plogp_v0”.
target (None) – unused since it is not a conditional generator.

Example

An example for using a generative algorithm from TorchDrug:

configuration = TorchDrugGCPN(algorithm_version=”qed_v0”) algorithm = TorchDrugGenerator(configuration=configuration) items = list(algorithm.sample(1)) print(items)

get_generator(configuration, target)[source]¶

Get the function to sample batches.

Parameters

configuration (AlgorithmConfiguration[~S, None]) – helps to set up the application.
target (None) – context or condition for the generation. Unused in the algorithm.

Return type

Callable[[], Iterable[Any]]

Returns

callable generating a batch of items.

validate_configuration(configuration)[source]¶

Overload to validate the a configuration for the algorithm.

Parameters: configuration (AlgorithmConfiguration) – the algorithm configuration.
Raises: InvalidAlgorithmConfiguration – in case the configuration for the algorithm is invalid.
Return type: AlgorithmConfiguration
Returns: the validated configuration.

__abstractmethods__ = frozenset({})¶

__annotations__ = {'generate': 'Untargeted', 'generator': 'Union[Untargeted, Targeted[T]]', 'max_runtime': 'int', 'max_samples': 'int', 'target': 'Optional[T]'}¶

__doc__ = None¶

__module__ = 'gt4sd.algorithms.generation.torchdrug.core'¶

__orig_bases__ = (gt4sd.algorithms.core.GeneratorAlgorithm[~S, NoneType],)¶

__parameters__ = (~S,)¶

_abc_impl = <_abc._abc_data object>¶

class TorchDrugGCPN(*args, **kwargs)[source]¶

Bases: TorchDrugGCPN, Generic[T]

Interface for TorchDrug Graph-convolutional policy network (GCPN) algorithm. Currently supported algorithm versions are “zinc250k_v0”, “qed_v0” and “plogp_v0”.

algorithm_type: ClassVar[str] = 'generation'¶: General type of generative algorithm.

domain: ClassVar[str] = 'materials'¶: General application domain. Hints at input/output types.

algorithm_version: str = 'zinc250k_v0'¶

To differentiate between different versions of an application.

There is no imposed naming convention.

get_conditional_generator(resources_path)[source]¶

Instantiate the actual generator implementation. :type resources_path: str :param resources_path: local path to model files.

Return type: GCPNGenerator
Returns: instance with sample method for generation.

classmethod get_filepath_mappings_for_training_pipeline_arguments()[source]¶

Get filepath mappings for the given training pipeline arguments. :type training_pipeline_arguments: TrainingPipelineArguments :param training_pipeline_arguments: training pipeline arguments.

Return type: Dict[str, str]
Returns: a mapping between artifacts’ files and training pipeline’s output files.

__annotations__ = {'algorithm_application': 'ClassVar[str]', 'algorithm_name': 'ClassVar[str]', 'algorithm_type': typing.ClassVar[str], 'algorithm_version': <class 'str'>, 'domain': typing.ClassVar[str]}¶

__dataclass_fields__ = {'algorithm_application': Field(name='algorithm_application',type=typing.ClassVar[str],default='TorchDrugGCPN',default_factory=<dataclasses._MISSING_TYPE object>,init=True,repr=True,hash=None,compare=True,metadata=mappingproxy({}),kw_only=<dataclasses._MISSING_TYPE object>,_field_type=_FIELD_CLASSVAR), 'algorithm_name': Field(name='algorithm_name',type=typing.ClassVar[str],default='TorchDrugGenerator',default_factory=<dataclasses._MISSING_TYPE object>,init=True,repr=True,hash=None,compare=True,metadata=mappingproxy({}),kw_only=<dataclasses._MISSING_TYPE object>,_field_type=_FIELD_CLASSVAR), 'algorithm_type': Field(name='algorithm_type',type=typing.ClassVar[str],default='generation',default_factory=<dataclasses._MISSING_TYPE object>,init=True,repr=True,hash=None,compare=True,metadata=mappingproxy({}),kw_only=<dataclasses._MISSING_TYPE object>,_field_type=_FIELD_CLASSVAR), 'algorithm_version': Field(name='algorithm_version',type=<class 'str'>,default='zinc250k_v0',default_factory=<dataclasses._MISSING_TYPE object>,init=True,repr=True,hash=None,compare=True,metadata=mappingproxy({}),kw_only=False,_field_type=_FIELD), 'domain': Field(name='domain',type=typing.ClassVar[str],default='materials',default_factory=<dataclasses._MISSING_TYPE object>,init=True,repr=True,hash=None,compare=True,metadata=mappingproxy({}),kw_only=<dataclasses._MISSING_TYPE object>,_field_type=_FIELD_CLASSVAR)}¶

__dataclass_params__ = _DataclassParams(init=True,repr=True,eq=True,order=False,unsafe_hash=False,frozen=False)¶

__doc__ = '\n Interface for TorchDrug Graph-convolutional policy network (GCPN) algorithm.\n Currently supported algorithm versions are "zinc250k_v0", "qed_v0" and "plogp_v0".\n '¶

__eq__(other)¶: Return self==value.

__hash__ = None¶

__init__(*args, **kwargs)¶

__match_args__ = ('algorithm_version',)¶

__module__ = 'gt4sd.algorithms.generation.torchdrug.core'¶

__orig_bases__ = (<class 'types.TorchDrugGCPN'>, typing.Generic[~T])¶

__parameters__ = (~T,)¶

__pydantic_complete__ = True¶

__pydantic_config__ = {}¶

__pydantic_core_schema__ = {'cls': <class 'gt4sd.algorithms.generation.torchdrug.core.TorchDrugGCPN'>, 'config': {'title': 'TorchDrugGCPN'}, 'fields': ['algorithm_version'], 'frozen': False, 'metadata': {'pydantic_js_annotation_functions': [], 'pydantic_js_functions': [functools.partial(<function modify_model_json_schema>, cls=<class 'gt4sd.algorithms.generation.torchdrug.core.TorchDrugGCPN'>, title=None)]}, 'post_init': False, 'ref': 'types.TorchDrugGCPN:94662758981984', 'schema': {'collect_init_only': False, 'computed_fields': [], 'dataclass_name': 'TorchDrugGCPN', 'fields': [{'type': 'dataclass-field', 'name': 'algorithm_version', 'schema': {'type': 'default', 'schema': {'type': 'str'}, 'default': 'zinc250k_v0'}, 'kw_only': False, 'init': True, 'metadata': {'pydantic_js_functions': [], 'pydantic_js_annotation_functions': [<function get_json_schema_update_func.<locals>.json_schema_update_func>]}}], 'type': 'dataclass-args'}, 'slots': True, 'type': 'dataclass'}¶

__pydantic_decorators__ = DecoratorInfos(validators={}, field_validators={}, root_validators={}, field_serializers={}, model_serializers={}, model_validators={}, computed_fields={})¶

__pydantic_fields__ = {'algorithm_version': FieldInfo(annotation=str, required=False, default='zinc250k_v0', init=True, init_var=False, kw_only=False)}¶

__pydantic_serializer__ = SchemaSerializer(serializer=Dataclass( DataclassSerializer { class: Py( 0x00005618641f6160, ), serializer: Fields( GeneralFieldsSerializer { fields: { "algorithm_version": SerField { key_py: Py( 0x00007f1dcae691b0, ), alias: None, alias_py: None, serializer: Some( WithDefault( WithDefaultSerializer { default: Default( Py( 0x00007f1dc39197f0, ), ), serializer: Str( StrSerializer, ), }, ), ), required: true, }, }, computed_fields: Some( ComputedFields( [], ), ), mode: SimpleDict, extra_serializer: None, filter: SchemaFilter { include: None, exclude: None, }, required_fields: 1, }, ), fields: [ Py( 0x00007f1ea52ed250, ), ], name: "TorchDrugGCPN", }, ), definitions=[])¶

__pydantic_validator__ = SchemaValidator(title="TorchDrugGCPN", validator=Dataclass( DataclassValidator { strict: false, validator: DataclassArgs( DataclassArgsValidator { fields: [ Field { kw_only: false, name: "algorithm_version", py_name: Py( 0x00007f1ea52ed250, ), init: true, init_only: false, lookup_key: Simple { key: "algorithm_version", py_key: Py( 0x00007f1dcae6b500, ), path: LookupPath( [ S( "algorithm_version", Py( 0x00007f1dcae69cf0, ), ), ], ), }, validator: WithDefault( WithDefaultValidator { default: Default( Py( 0x00007f1dc39197f0, ), ), on_error: Raise, validator: Str( StrValidator { strict: false, coerce_numbers_to_str: false, }, ), validate_default: false, copy_default: false, name: "default[str]", undefined: Py( 0x00007f1ea71db950, ), }, ), frozen: false, }, ], positional_count: 1, init_only_count: None, dataclass_name: "TorchDrugGCPN", validator_name: "dataclass-args[TorchDrugGCPN]", extra_behavior: Ignore, extras_validator: None, loc_by_alias: true, }, ), class: Py( 0x00005618641f6160, ), fields: [ Py( 0x00007f1ea52ed250, ), ], post_init: None, revalidate: Never, name: "TorchDrugGCPN", frozen: false, slots: true, }, ), definitions=[], cache_strings=True)¶

__repr__()¶: Return repr(self).

__signature__ = <Signature (algorithm_version: str = 'zinc250k_v0') -> None>¶

__wrapped__¶: alias of TorchDrugGCPN

class TorchDrugGraphAF(*args, **kwargs)[source]¶

Bases: TorchDrugGraphAF, Generic[T]

Interface for TorchDrug flow-based autoregressive graph algorithm (GraphAF). Currently supported algorithm versions are “zinc250k_v0”, “qed_v0” and “plogp_v0”.

algorithm_type: ClassVar[str] = 'generation'¶: General type of generative algorithm.

domain: ClassVar[str] = 'materials'¶: General application domain. Hints at input/output types.