Models

libcll supports 4 commonly-used deep learning models, as shown in the table below.

Models	Description
`Linear`	Single-layer neural network (input_dim-num_classes) for simple datasets.
`MLP`	Multi-layer neural network (input_dim-hidden_dim-num_classes) for simple datasets.
`ResNet`	Residual neural network for hard datasets.
`DenseNet`	Densely connected convolutional network for hard datasets.

Loading a Model

To build a model, libcll requires input dimension and the number of classes from dataset and hidden dimension if needed.

from libcll.models import build_model

model = build_model(
    model="MLP",
    input_dim=input_dim,
    hidden_dim=512,
    num_classes=num_classes,
)