logitorch.models.fairr

Classes

FactSelector	FactSelector is a class that represents a fact-based selector model.
FaiRR	FaiRR is a class that represents the FaiRR model, which combines rule-based and fact-based selectors.
KnowledgeComposer	KnowledgeComposer is a class that represents a knowledge composer model.
RuleSelector	RuleSelector is a class that represents a rule-based selector model.

Module Contents

class logitorch.models.fairr.FactSelector(pretrained_roberta_model: str)

Bases: torch.nn.Module

FactSelector is a class that represents a fact-based selector model. It uses a pretrained RoBERTa model for encoding input sequences and a linear classifier for prediction.

forward(x, y=None)

Forward pass of the FactSelector model.

Args:

x (dict): Input dictionary containing the input sequences. y (None): Placeholder for compatibility with other models.

Returns:

None: This method is not implemented.

classifier

dropout

model

out_dim

tokenizer

class logitorch.models.fairr.FaiRR

Bases: torch.nn.Module

FaiRR is a class that represents the FaiRR model, which combines rule-based and fact-based selectors.

class logitorch.models.fairr.KnowledgeComposer(pretrained_t5_model: str)

Bases: torch.nn.Module

KnowledgeComposer is a class that represents a knowledge composer model. It uses a pretrained T5 model for generating text based on input prompts.

model

tokenizer

class logitorch.models.fairr.RuleSelector(pretrained_roberta_model: str, cls_dropout=0.1)

Bases: torch.nn.Module

RuleSelector is a class that represents a rule-based selector model. It uses a pretrained RoBERTa model for encoding input sequences and a linear classifier for prediction.

forward(x, y=None)

Forward pass of the RuleSelector model.

Args:

x (dict): Input dictionary containing the input sequences. y (None): Placeholder for compatibility with other models.

Returns:

torch.Tensor: Logits representing the predicted scores.

classifier

dropout

model

out_dim

tokenizer