reagent.net_builder.slate_ranking package

Submodules

reagent.net_builder.slate_ranking.slate_ranking_scorer module

class reagent.net_builder.slate_ranking.slate_ranking_scorer.FinalLayer(score_cap: Optional[float] = None, sigmoid: bool = False, tanh: bool = False)

Bases: object

get()

score_cap: Optional[float] = None

sigmoid: bool = False

tanh: bool = False

class reagent.net_builder.slate_ranking.slate_ranking_scorer.ScoreCap(cap: float)

Bases: torch.nn.modules.module.Module

forward(input)

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.

training: bool

class reagent.net_builder.slate_ranking.slate_ranking_scorer.SlateRankingScorer(hidden_layers: List[int] = <factory>, activations: List[str] = <factory>, use_batch_norm: bool = False, min_std: float = 0.0, dropout_ratio: float = 0.0, use_layer_norm: bool = False, normalize_output: bool = False, orthogonal_init: bool = False, has_user_feat: bool = False, final_layer: reagent.net_builder.slate_ranking.slate_ranking_scorer.FinalLayer = <factory>)

Bases: reagent.net_builder.slate_ranking_net_builder.SlateRankingNetBuilder

activations: List[str]

build_slate_ranking_network(state_dim, candidate_dim, _candidate_size=None, _slate_size=None) → reagent.models.base.ModelBase

dropout_ratio: float = 0.0

final_layer: reagent.net_builder.slate_ranking.slate_ranking_scorer.FinalLayer

has_user_feat: bool = False

hidden_layers: List[int]

min_std: float = 0.0

normalize_output: bool = False

orthogonal_init: bool = False

use_batch_norm: bool = False

use_layer_norm: bool = False

reagent.net_builder.slate_ranking.slate_ranking_transformer module

class reagent.net_builder.slate_ranking.slate_ranking_transformer.SlateRankingTransformer(output_arch: reagent.model_utils.seq2slate_utils.Seq2SlateOutputArch = <Seq2SlateOutputArch.AUTOREGRESSIVE: 'autoregressive'>, temperature: float = 1.0, transformer: reagent.core.parameters.TransformerParameters = <factory>)

Bases: reagent.net_builder.slate_ranking_net_builder.SlateRankingNetBuilder

build_slate_ranking_network(state_dim, candidate_dim, candidate_size, slate_size) → reagent.models.base.ModelBase

output_arch: reagent.model_utils.seq2slate_utils.Seq2SlateOutputArch = 'autoregressive'

temperature: float = 1.0

transformer: reagent.core.parameters.TransformerParameters

Module contents

class reagent.net_builder.slate_ranking.SlateRankingNetBuilder__Union(SlateRankingTransformer: Optional[reagent.net_builder.slate_ranking.slate_ranking_transformer.SlateRankingTransformer] = None, SlateRankingScorer: Optional[reagent.net_builder.slate_ranking.slate_ranking_scorer.SlateRankingScorer] = None)

Bases: reagent.core.tagged_union.TaggedUnion

SlateRankingScorer: Optional[reagent.net_builder.slate_ranking.slate_ranking_scorer.SlateRankingScorer] = None

SlateRankingTransformer: Optional[reagent.net_builder.slate_ranking.slate_ranking_transformer.SlateRankingTransformer] = None