ResNet¶

class fortuna.model.resnet.ResNet(stage_sizes, block_cls, output_dim, num_filters=64, dtype=<class 'jax.numpy.float32'>, activation=<jax._src.custom_derivatives.custom_jvp object>, conv=<class 'flax.linen.linear.Conv'>, parent=<flax.linen.module._Sentinel object>, name=None)[source]¶

Bases: Module

Deep feature extractor subnetwork.

stage_sizes¶

Sizes for each stage.

Type:: Sequence[int]

block_cls¶

Block class.

Type:: ModuleDef

output_dim¶

Output dimension.

Type:: int

num_filters¶

Number of filters.

Type:: int

dtype¶

Layers’ dtype.

Type:: Any

activation¶

Activation function.

Type:: Callable

conv¶

Convolution module.

Type:: ModuleDef

fortuna.model.resnet.ResNet18¶: alias of functools.partial(<class ‘fortuna.model.resnet.ResNet’>, stage_sizes=[2, 2, 2, 2], block_cls=<class ‘fortuna.model.resnet.ResNetBlock’>)

fortuna.model.resnet.ResNet34¶: alias of functools.partial(<class ‘fortuna.model.resnet.ResNet’>, stage_sizes=[3, 4, 6, 3], block_cls=<class ‘fortuna.model.resnet.ResNetBlock’>)

fortuna.model.resnet.ResNet50¶: alias of functools.partial(<class ‘fortuna.model.resnet.ResNet’>, stage_sizes=[3, 4, 6, 3], block_cls=<class ‘fortuna.model.resnet.BottleneckResNetBlock’>)

fortuna.model.resnet.ResNet101¶: alias of functools.partial(<class ‘fortuna.model.resnet.ResNet’>, stage_sizes=[3, 4, 23, 3], block_cls=<class ‘fortuna.model.resnet.BottleneckResNetBlock’>)

fortuna.model.resnet.ResNet152¶: alias of functools.partial(<class ‘fortuna.model.resnet.ResNet’>, stage_sizes=[3, 8, 36, 3], block_cls=<class ‘fortuna.model.resnet.BottleneckResNetBlock’>)

fortuna.model.resnet.ResNet200¶: alias of functools.partial(<class ‘fortuna.model.resnet.ResNet’>, stage_sizes=[3, 24, 36, 3], block_cls=<class ‘fortuna.model.resnet.BottleneckResNetBlock’>)