carate.models package

Submodules

carate.models.base_model module

class carate.models.base_model.Model(dim: int, num_classes: int, num_features: int)

Bases: Module

abstract forward(x: int, edge_index: int, batch: int, edge_weight=None) → Tensor

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

carate.models.cc_classification module

cgc_classification model is named after the structure of the graph neural network. The graph neural network is structured with a convolutional , graph attention, and another convolutional layer. The cgc_classificatin model was the model tested int the publication Introducing CARATE: Finally speaking chemistry.

author:

Julian M. Kleber

class carate.models.cc_classification.Net(dim: int, num_features: int, num_classes: int, *args, **kwargs)

Bases: Model

The Net is the core algorithm and needs a constructor and a forward pass. The train, test and evaluation methods are implemented in the evaluation module with the Evaluation class.

forward(x, edge_index, batch, edge_weight=None)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

carate.models.cc_regression module

cc_regression model is named after the structure of the graph neural network. The graph neural network is structured with a convolutional, and another convolutional layer. The cgc_classificatin model was the model tested int the publication Introducing CARATE: Finally speaking chemistry.

class carate.models.cc_regression.Net(dim: int, num_features: int, num_classes: int, *args, **kwargs)

Bases: Model

forward(x: float, edge_index: int, batch: int, edge_weight=None) → float

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

carate.models.cgc_classification module

cgc_classification model is named after the structure of the graph neural network. The graph neural network is structured with a convolutional , graph attention, and another convolutional layer. The cgc_classificatin model was the model tested int the publication Introducing CARATE: Finally speaking chemistry.

author:

Julian M. Kleber

class carate.models.cgc_classification.Net(dim: int, num_features: int, num_classes: int, num_heads: int = 16, dropout_gat: float = 0.6, dropout_forward=0.5, *args, **kwargs)

Bases: Model

The Net is the core algorithm and needs a constructor and a forward pass. The train, test and evaluation methods are implemented in the evaluation module with the Evaluation class.

forward(x, edge_index, batch, edge_weight=None)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

carate.models.cgc_classification_jax module

carate.models.cgc_regression module

cgc_regression model is named after the structure of the graph neural network. The graph neural network is structured with a convolutional , graph attention, and another convolutional layer. The cgc_classificatin model was the model tested int the publication Introducing CARATE: Finally speaking chemistry.

class carate.models.cgc_regression.Net(dim: int, num_features: int, num_classes: int, num_heads: int = 16, dropout_gat: float = 0.6, dropout_forward=0.5, *args, **kwargs)

Bases: Model

forward(x: float, edge_index: int, batch: int, edge_weight=None) → float

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

carate.models.cgc_regression_jax module

carate.models.g_classification module

class carate.models.g_classification.Net(dim: int, num_features: int, num_classes: int, num_heads: int = 16, dropout_gat: float = 0.6, dropout_forward=0.5, *args, **kwargs)

Bases: Model

The Net is the core algorithm and needs a constructor and a forward pass. The train, test and evaluation methods are implemented in the evaluation module with the Evaluation class.

forward(x, edge_index, batch, edge_weight=None)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

carate.models.g_regression module

class carate.models.g_regression.Net(dim: int, num_features: int, num_classes: int, num_heads: int = 16, dropout_gat: float = 0.6, dropout_forward: float = 0.5, *args, **kwargs)

Bases: Model

The Net is the core algorithm and needs a constructor and a forward pass. The train, test and evaluation methods are implemented in the evaluation module with the Evaluation class.

forward(x, edge_index, batch, edge_weight=None)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

carate.models.gatv2_classification module

class carate.models.gatv2_classification.Net(dim: int, num_features: int, num_classes: int, heads: int = 16, dropout_gat: float = 0.6, dropout_forward=0.5, *args, **kwargs)

Bases: Model

The Net is the core algorithm and needs a constructor and a forward pass. The train, test and evaluation methods are implemented in the evaluation module with the Evaluation class.

forward(x, edge_index, batch, edge_weight=None)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

carate.models.gatv2_regression module

class carate.models.gatv2_regression.Net(dim: int, num_features: int, num_classes: int, num_heads: int = 16, dropout_gat: float = 0.6, dropout_forward=0.5, *args, **kwargs)

Bases: Model

The Net is the core algorithm and needs a constructor and a forward pass. The train, test and evaluation methods are implemented in the evaluation module with the Evaluation class.

forward(x, edge_index, batch, edge_weight=None)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

carate.models.gcc_classification module

cgc_classification model is named after the structure of the graph neural network. The graph neural network is structured with a convolutional , graph attention, and another convolutional layer. The cgc_classificatin model was the model tested int the publication Introducing CARATE: Finally speaking chemistry.

author:

Julian M. Kleber

class carate.models.gcc_classification.Net(dim: int, num_features: int, num_classes: int, num_heads: int = 16, dropout_gat: float = 0.6, dropout_forward=0.5, *args, **kwargs)

Bases: Model

The Net is the core algorithm and needs a constructor and a forward pass. The train, test and evaluation methods are implemented in the evaluation module with the Evaluation class.

forward(x, edge_index, batch, edge_weight=None)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

carate.models.gcc_regression module

cgc_classification model is named after the structure of the graph neural network. The graph neural network is structured with a convolutional , graph attention, and another convolutional layer. The cgc_classificatin model was the model tested int the publication Introducing CARATE: Finally speaking chemistry.

author:

Julian M. Kleber

class carate.models.gcc_regression.Net(dim: int, num_features: int, num_classes: int, num_heads: int = 16, dropout_gat: float = 0.6, dropout_forward=0.5, *args, **kwargs)

Bases: Model

The Net is the core algorithm and needs a constructor and a forward pass. The train, test and evaluation methods are implemented in the evaluation module with the Evaluation class.

forward(x, edge_index, batch, edge_weight=None)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

carate.models.graph_transformer_classification module

class carate.models.graph_transformer_classification.Net(dim: int, num_features: int, num_classes: int, num_heads: int = 16, dropout_gat: float = 0.6, dropout_forward: float = 0.5, *args, **kwargs)

Bases: Model

The Net is the core algorithm and needs a constructor and a forward pass. The train, test and evaluation methods are implemented in the evaluation module with the Evaluation class.

forward(x, edge_index, batch, edge_weight=None)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

carate.models.graph_transformer_regression module

class carate.models.graph_transformer_regression.Net(dim: int, num_features: int, num_classes: int, num_heads: int = 16, dropout_gat: float = 0.6, dropout_forward: float = 0.5, *args, **kwargs)

Bases: Model

The Net is the core algorithm and needs a constructor and a forward pass. The train, test and evaluation methods are implemented in the evaluation module with the Evaluation class. The net takes plain graph transformer and applies common postprocessing.

forward(x, edge_index, batch, edge_weight=None)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

Module contents