ISubGVQA.sampling.methods.simple_scheme

Attributes

LARGE_NUMBER

Classes

EdgeSIMPLEBatched

Functions

logsigmoid(x)

Module Contents

ISubGVQA.sampling.methods.simple_scheme.LARGE_NUMBER = 10000000000.0

ISubGVQA.sampling.methods.simple_scheme.logsigmoid(x)

class ISubGVQA.sampling.methods.simple_scheme.EdgeSIMPLEBatched(k, device, policy, val_ensemble=1, train_ensemble=1, logits_activation=None)

Bases: torch.nn.Module

k

device

policy

layer_configs

adj = None

val_ensemble = 1

train_ensemble = 1

logits_activation = None

forward(scores, train=True)

validation(scores)

during the inference we need to margin-out the stochasticity thus we do top-k once or sample multiple times

Args:: scores: shape B x N x N x E
Returns:: mask: shape B x N x N x (E x VE)