ISubGVQA.sampling.methods.gumbel_scheme

Attributes

`EPSILON`
`LARGE_NUMBER`

Classes

GumbelSampler

Module Contents

ISubGVQA.sampling.methods.gumbel_scheme.EPSILON

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

class ISubGVQA.sampling.methods.gumbel_scheme.GumbelSampler(k, train_ensemble, val_ensemble, tau=0.1, hard=True, policy=None)

Bases: torch.nn.Module

policy = None

k

hard = True

tau = 0.1

adj = None

train_ensemble

val_ensemble

forward(scores, train=True)

validation(scores)