sakura.models.modules.modulebuilder

sakura.models.modules.modulebuilder(cfg)

Builds neural network layers in sequential order based on configurations.

Module configuration:

Each dict should contain key ‘type’ (str): Module type identifier. Supported types:

• ‘Linear’: Requires ‘in_dim’ and ‘out_dim’ keys

• ‘Dropout’: Optional ‘p’ key (default: 0.5)

• ‘ReLU’, ‘CELU’, ‘Softmax’, ‘LogSoftmax’: No additional params

Parameters:

cfg (list[dict]) – A list of module configuration dictionaries

Returns:

Ordered list of initialized PyTorch modules

Return type:

list[nn.Module]