Skip to content

the parameters in default.ini #92

@Dengsongquanch

Description

@Dengsongquanch

There are many parameters in default.ini that are not explained in document:
[train]
early_stopping_loss = 0.0
early_stopping_loss_epoch = [0.000000, 500]
revert_then_decay = True
revert_threshold = 30
revert_decay_epoch = [500, 2000, 3000]
revert_decay_gamma = [0.4, 0.5, 0.5]
clip_grad = True
clip_grad_value = 4.2
switch_sgd = False
switch_sgd_lr = 1e-4
switch_sgd_epoch = -1
[hyperparameter]
lr_scheduler =
lr_milestones = []
momentum = 0.9
weight_decay = 0
criterion = MaskMSELoss
retain_edge_fea = True
lambda_Eij = 0.0
lambda_Ei = 0.1
lambda_Etot = 0.0
[network]
aggr = add
if_exp = True
if_MultipleLinear = False
if_edge_update = True
if_lcmp = True
trainable_gaussians = False
type_affine = False

where can I find the explanation of these parameters?

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions