nn.Dense

class nn.Dense(*args: Any, **kwargs: Any)

Fully connected linear layer with activation function.

\[y = activation(x W^T + b)\]

Parameters:

• in_features – number of input feature \(x\).

• out_features – umber of output features \(y\).

• bias – If False, the layer will not adapt bias \(b\).

• activation – if None, no activation function is used.

• weight_init – weight initializer from current weight.

• bias_init – bias initializer from current bias.