model.GCN package

Submodules

model.GCN.GCN module

class model.GCN.GCN.GCNResnet(model, num_classes, in_channel=300, t=0, adj_file=None)

Bases: torch.nn.modules.module.Module

forward(feature, inp)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

get_config_optim(lr, lrp)
class model.GCN.GCN.GraphAttentionLayer(in_features, out_features, dropout, alpha, batch_size, concat=True)

Bases: torch.nn.modules.module.Module

Graph Attention Layer. Reference https://arxiv.org/abs/1710.10903 the basic GCN can be modified to this layer in recognition code

forward(input, adj)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

reset_parameters()
class model.GCN.GCN.GraphConvolution(in_features, out_features, dropout=0.0, bias=True)

Bases: torch.nn.modules.module.Module

Simple GCN layer, similar to https://arxiv.org/abs/1609.02907

in_features: input feature dimension out_features: output feature dimension

dropout: default=0 not used， call nn.dropout function

bias: the learnable bias in GCN layer, default set as True

forward(input, adj)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

reset_parameters()
model.GCN.GCN.gen_A(num_classes, t, adj_file)
model.GCN.GCN.norm_adj(A)
model.GCN.GCN.norm_adj_batch(A)

model.GCN.Graph module

model.GCN.Graphsage module

class model.GCN.Graphsage.BatchedGraphSAGE(infeat, outfeat, use_bn=False, mean=False, add_self=False)

Bases: torch.nn.modules.module.Module

forward(x, adj, mask=None)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

model.GCN.non_local module

class model.GCN.non_local.Non_local(in_channels, reduc_ratio=2)

Bases: torch.nn.modules.module.Module

forward(x)
Parameters

x -- (b, t, h, w)

Return x

(b, t, h, w)

model.GCN.sparseGCN module

class model.GCN.sparseGCN.SpGraphAttentionLayer(in_features, out_features, dropout, alpha, concat=True)

Bases: torch.nn.modules.module.Module

Sparse version GAT layer, similar to https://arxiv.org/abs/1710.10903

forward(input, adj)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

class model.GCN.sparseGCN.SpecialSpmm

Bases: torch.nn.modules.module.Module

forward(indices, values, shape, b)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

class model.GCN.sparseGCN.SpecialSpmmFunction

Bases: torch.autograd.function.Function

Special function for only sparse region backpropataion layer.

static backward(ctx, grad_output)

Defines a formula for differentiating the operation.

This function is to be overridden by all subclasses.

It must accept a context ctx as the first argument, followed by as many outputs did forward() return, and it should return as many tensors, as there were inputs to forward(). Each argument is the gradient w.r.t the given output, and each returned value should be the gradient w.r.t. the corresponding input.

The context can be used to retrieve tensors saved during the forward pass. It also has an attribute ctx.needs_input_grad as a tuple of booleans representing whether each input needs gradient. E.g., backward() will have ctx.needs_input_grad[0] = True if the first input to forward() needs gradient computated w.r.t. the output.

static forward(ctx, indices, values, shape, b)

Performs the operation.

This function is to be overridden by all subclasses.

It must accept a context ctx as the first argument, followed by any number of arguments (tensors or other types).

The context can be used to store tensors that can be then retrieved during the backward pass.

Module contents