MolecularDiffusion.modules.layers.equiformer_v2.drop¶

Add extra_repr into DropPath implemented by timm for displaying more info.

Classes¶

`DropPath`	Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks).
`EquivariantDropout`
`EquivariantDropoutArraySphericalHarmonics`
`EquivariantScalarsDropout`
`GraphDropPath`	Consider batch for graph data when dropping paths.

Functions¶

drop_path(x[, drop_prob, training])

Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks).

Module Contents¶

class MolecularDiffusion.modules.layers.equiformer_v2.drop.DropPath(drop_prob=None)¶

Bases: torch.nn.Module

Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks).

extra_repr()¶

forward(x)¶

drop_prob = None¶

class MolecularDiffusion.modules.layers.equiformer_v2.drop.EquivariantDropout(irreps, drop_prob)¶

Bases: torch.nn.Module

forward(x)¶

drop¶

drop_prob¶

irreps¶

mul¶

num_irreps¶

class MolecularDiffusion.modules.layers.equiformer_v2.drop.EquivariantDropoutArraySphericalHarmonics(drop_prob, drop_graph=False)¶

Bases: torch.nn.Module

extra_repr()¶

forward(x, batch=None)¶

drop¶

drop_graph = False¶

drop_prob¶

class MolecularDiffusion.modules.layers.equiformer_v2.drop.EquivariantScalarsDropout(irreps, drop_prob)¶

Bases: torch.nn.Module

extra_repr()¶

forward(x)¶

drop_prob¶

irreps¶

class MolecularDiffusion.modules.layers.equiformer_v2.drop.GraphDropPath(drop_prob=None)¶

Bases: torch.nn.Module

Consider batch for graph data when dropping paths.

extra_repr()¶

forward(x, batch)¶

drop_prob = None¶

MolecularDiffusion.modules.layers.equiformer_v2.drop.drop_path(x, drop_prob: float = 0.0, training: bool = False)¶

Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks).

This is the same as the DropConnect impl I created for EfficientNet, etc networks, however, the original name is misleading as ‘Drop Connect’ is a different form of dropout in a separate paper… See discussion: https://github.com/tensorflow/tpu/issues/494#issuecomment-532968956 … I’ve opted for changing the layer and argument names to ‘drop path’ rather than mix DropConnect as a layer name and use ‘survival rate’ as the argument.