unitorch.cli.models

ClassificationOutputs

Bases: TensorsOutputs, WriterMixin

Outputs for classification models.

Parameters:

Name	Type	Description	Default
outputs	Tensor	Output tensor containing the classification results.	required

ClassificationTargets

Bases: TensorsTargets

Targets for classification models.

Parameters:

Name	Type	Description	Default
targets	Tensor	The target tensor.	required
sample_weight	Optional[Tensor]	The weight associated with each target. Defaults to a tensor with a value of 1.0.	tensor(1.0)

EmbeddingOutputs

Bases: TensorsOutputs, WriterMixin

Outputs for embedding models.

Parameters:

Name	Type	Description	Default
embedding	Tensor	The embedding tensor.	required
embedding1	Optional[Tensor]	Additional embedding tensor 1. Defaults to an empty tensor.	empty(0)
embedding2	Optional[Tensor]	Additional embedding tensor 2. Defaults to an empty tensor.	empty(0)
embedding3	Optional[Tensor]	Additional embedding tensor 3. Defaults to an empty tensor.	empty(0)
embedding4	Optional[Tensor]	Additional embedding tensor 4. Defaults to an empty tensor.	empty(0)

GenerationOutputs

Bases: TensorsOutputs, WriterMixin

Outputs for generation models.

Parameters:

Name	Type	Description	Default
sequences	Tensor	Generated sequences.	required
sequences_scores	Optional[Tensor]	Scores associated with the generated sequences. Defaults to an empty tensor.	empty(0)

GenerationTargets

Bases: TensorsTargets

Targets for generation models.

Parameters:

Name	Type	Description	Default
refs	Tensor	Reference sequences.	required
masks	Optional[Tensor]	Mask indicating the valid positions in the reference sequences. Defaults to an empty tensor.	empty(0)
sample_weight	Optional[Tensor]	Sample weights for the reference sequences. Defaults to an empty tensor.	empty(0)

LossOutputs

Bases: TensorsOutputs