MLPDecoder.Config¶

Component: MLPDecoder

class MLPDecoder.Config[source]

Bases: DecoderBase.Config

Configuration class for MLPDecoder.

hidden_dims¶

Dimensions of the outputs of hidden layers..

Type:	List[int]

temperature¶

Scales logits by this value (before the softmax

Type:	float

operation) during test-time only. Temperature scaling has no effect on

the top prediction but changes the shape of the posterior distribution,

which can be useful for a range of tasks

Type:	e.g., model calibration

All Attributes (including base classes)

load_path: Optional[str] = None

save_path: Optional[str] = None

freeze: bool = False

shared_module_key: Optional[str] = None

hidden_dims: list[int] = []

out_dim: Optional[int] = None

layer_norm: bool = False

dropout: float = 0.0

bias: bool = True

activation: Activation = <Activation.RELU: 'relu'>

temperature: float = 1.0

spectral_normalization: bool = False

Default JSON

{
    "load_path": null,
    "save_path": null,
    "freeze": false,
    "shared_module_key": null,
    "hidden_dims": [],
    "out_dim": null,
    "layer_norm": false,
    "dropout": 0.0,
    "bias": true,
    "activation": "relu",
    "temperature": 1.0,
    "spectral_normalization": false
}