def arg_parse():
parser = argparse.ArgumentParser(description="GraphPool arguments.")
io_parser = parser.add_mutually_exclusive_group(required=False)
io_parser.add_argument("--dataset", dest="dataset", help="Input dataset.")
benchmark_parser = io_parser.add_argument_group()
benchmark_parser.add_argument("--bmname",
dest="bmname",
help="Name of the benchmark dataset")
io_parser.add_argument("--pkl",
dest="pkl_fname",
help="Name of the pkl data file")
softpool_parser = parser.add_argument_group()
softpool_parser.add_argument(
"--assign-ratio",
dest="assign_ratio",
type=float,
help="ratio of number of nodes in consecutive layers",
)
softpool_parser.add_argument("--num-pool",
dest="num_pool",
type=int,
help="number of pooling layers")
parser.add_argument(
"--linkpred",
dest="linkpred",
action="store_const",
const=True,
default=False,
help="Whether link prediction side objective is used",
)
parser_utils.parse_optimizer(parser)
parser.add_argument("--datadir",
dest="datadir",
help="Directory where benchmark is located")
parser.add_argument("--logdir",
dest="logdir",
help="Tensorboard log directory")
parser.add_argument("--ckptdir",
dest="ckptdir",
help="Model checkpoint directory")
parser.add_argument("--cuda", dest="cuda", help="CUDA.")
parser.add_argument(
"--gpu",
dest="gpu",
action="store_const",
const=True,
default=False,
help="whether to use GPU.",
)
parser.add_argument(
"--max-nodes",
dest="max_nodes",
type=int,
help=
"Maximum number of nodes (ignore graghs with nodes exceeding the number.",
)
parser.add_argument("--batch-size",
dest="batch_size",
type=int,
help="Batch size.")
parser.add_argument("--epochs",
dest="num_epochs",
type=int,
help="Number of epochs to train.")
parser.add_argument(
"--train-ratio",
dest="train_ratio",
type=float,
help="Ratio of number of graphs training set to all graphs.",
)
parser.add_argument(
"--num_workers",
dest="num_workers",
type=int,
help="Number of workers to load data.",
)
parser.add_argument(
"--feature",
dest="feature_type",
help="Feature used for encoder. Can be: id, deg",
)
parser.add_argument("--input-dim",
dest="input_dim",
type=int,
help="Input feature dimension")
parser.add_argument("--hidden-dim",
dest="hidden_dim",
type=int,
help="Hidden dimension")
parser.add_argument("--output-dim",
dest="output_dim",
type=int,
help="Output dimension")
parser.add_argument("--num-classes",
dest="num_classes",
type=int,
help="Number of label classes")
parser.add_argument(
"--num-gc-layers",
dest="num_gc_layers",
type=int,
help="Number of graph convolution layers before each pooling",
)
parser.add_argument(
"--bn",
dest="bn",
action="store_const",
const=True,
default=False,
help="Whether batch normalization is used",
)
parser.add_argument("--dropout",
dest="dropout",
type=float,
help="Dropout rate.")
parser.add_argument(
"--nobias",
dest="bias",
action="store_const",
const=False,
default=True,
help="Whether to add bias. Default to True.",
)
parser.add_argument(
"--weight-decay",
dest="weight_decay",
type=float,
help="Weight decay regularization constant.",
)
parser.add_argument("--method",
dest="method",
help="Method. Possible values: base, ")
parser.add_argument("--name-suffix",
dest="name_suffix",
help="suffix added to the output filename")
parser.set_defaults(
datadir="data", # io_parser
logdir="log",
ckptdir="ckpt",
dataset="syn1",
opt="adam", # opt_parser
opt_scheduler="none",
max_nodes=100,
cuda="1",
feature_type="default",
lr=0.001,
clip=2.0,
batch_size=20,
num_epochs=1000,
train_ratio=0.8,
test_ratio=0.1,
num_workers=1,
input_dim=10,
hidden_dim=20,
output_dim=20,
num_classes=2,
num_gc_layers=3,
dropout=0.0,
weight_decay=0.005,
method="base",
name_suffix="",
assign_ratio=0.1,
)
return parser.parse_args()
def arg_parse():
parser = argparse.ArgumentParser(description="GNN Explainer arguments.")
io_parser = parser.add_mutually_exclusive_group(required=False)
io_parser.add_argument("--dataset", dest="dataset", help="Input dataset.")
benchmark_parser = io_parser.add_argument_group()
benchmark_parser.add_argument("--bmname",
dest="bmname",
help="Name of the benchmark dataset")
io_parser.add_argument("--pkl",
dest="pkl_fname",
help="Name of the pkl data file")
parser_utils.parse_optimizer(parser)
parser.add_argument(
"--clean-log",
action="store_true",
help="If true, cleans the specified log directory before running.")
parser.add_argument("--logdir",
dest="logdir",
help="Tensorboard log directory")
parser.add_argument("--ckptdir",
dest="ckptdir",
help="Model checkpoint directory")
parser.add_argument("--cuda", dest="cuda", help="CUDA.")
parser.add_argument(
"--gpu",
dest="gpu",
action="store_const",
const=True,
default=False,
help="whether to use GPU.",
)
parser.add_argument("--epochs",
dest="num_epochs",
type=int,
help="Number of epochs to train.")
parser.add_argument("--hidden-dim",
dest="hidden_dim",
type=int,
help="Hidden dimension")
parser.add_argument("--output-dim",
dest="output_dim",
type=int,
help="Output dimension")
parser.add_argument(
"--num-gc-layers",
dest="num_gc_layers",
type=int,
help="Number of graph convolution layers before each pooling",
)
parser.add_argument(
"--bn",
dest="bn",
action="store_const",
const=True,
default=False,
help="Whether batch normalization is used",
)
parser.add_argument("--dropout",
dest="dropout",
type=float,
help="Dropout rate.")
parser.add_argument(
"--nobias",
dest="bias",
action="store_const",
const=False,
default=True,
help="Whether to add bias. Default to True.",
)
parser.add_argument(
"--no-writer",
dest="writer",
action="store_const",
const=False,
default=True,
help="Whether to add bias. Default to True.",
)
# Explainer
parser.add_argument("--mask-act",
dest="mask_act",
type=str,
help="sigmoid, ReLU.")
parser.add_argument(
"--mask-bias",
dest="mask_bias",
action="store_const",
const=True,
default=False,
help="Whether to add bias. Default to True.",
)
parser.add_argument("--explain-node",
dest="explain_node",
type=int,
help="Node to explain.")
parser.add_argument("--graph-idx",
dest="graph_idx",
type=int,
help="Graph to explain.")
parser.add_argument(
"--graph-mode",
dest="graph_mode",
action="store_const",
const=True,
default=False,
help="whether to run Explainer on Graph Classification task.",
)
parser.add_argument(
"--multigraph-class",
dest="multigraph_class",
type=int,
help=
"whether to run Explainer on multiple Graphs from the Classification task for examples in the same class.",
)
parser.add_argument(
"--multinode-class",
dest="multinode_class",
type=int,
help=
"whether to run Explainer on multiple nodes from the Classification task for examples in the same class.",
)
parser.add_argument(
"--align-steps",
dest="align_steps",
type=int,
help="Number of iterations to find P, the alignment matrix.",
)
parser.add_argument("--method",
dest="method",
type=str,
help="Method. Possible values: base, att.")
parser.add_argument("--name-suffix",
dest="name_suffix",
help="suffix added to the output filename")
parser.add_argument(
"--explainer-suffix",
dest="explainer_suffix",
help="suffix added to the explainer log",
)
parser.add_argument(
"--hops",
dest="hops",
type=int,
help="k-hop subgraph considered for GraphSHAP",
)
parser.add_argument(
"--num_samples",
dest="num_samples",
type=int,
help="number of samples used to train GraphSHAP",
)
parser.add_argument(
"--multiclass",
type=bool,
help='False if we consider explanations for the predicted class only')
parser.add_argument(
"--hv",
type=str,
help="way simplified input is translated to the original input space")
parser.add_argument("--feat",
type=str,
help="node features considered for hv above")
parser.add_argument("--coal", type=str, help="type of coalition sampler")
parser.add_argument("--g",
type=str,
help="method used to train g on derived dataset")
parser.add_argument(
"--regu",
type=int,
help='None if we do not apply regularisation, 1 if only feat')
parser.add_argument("--info",
type=bool,
help='True if we want to see info about the explainer')
parser.add_argument(
"--fullempty",
type=str,
default=None,
help='True if want to discard full and empty coalitions')
parser.add_argument(
"--S",
type=int,
default=3,
help=
'Max size of coalitions sampled in priority and treated specifically')
# TODO: Check argument usage
parser.set_defaults(
logdir="log",
ckptdir="ckpt",
dataset="syn1",
#bmname='Mutagenicity',
graph_mode=False,
opt="adam",
opt_scheduler="none",
gpu=False,
cuda="0",
lr=0.1,
clip=2.0,
batch_size=20,
num_epochs=150,
hidden_dim=20,
output_dim=20,
num_gc_layers=3,
dropout=0.0,
method="base",
name_suffix="",
explainer_suffix="",
align_steps=100,
explain_node=None,
graph_idx=-1,
mask_act="sigmoid",
multigraph_class=-1,
multinode_class=-1,
hops=2,
num_samples=65,
multiclass=False,
fullempty=None,
S=1,
hv='compute_pred',
feat='Expectation',
coal='SmarterSeparate',
g='WLR_sklearn',
regu=None,
info=True,
)
return parser.parse_args()
def arg_parse():
print("Attempt to parse arguments...")
parser = argparse.ArgumentParser(description='GNN program arguments.')
# Set optimizer parameters
parser_utils.parse_optimizer(parser)
## Add parsing arguments
parser.add_argument('--datadir',
dest='datadir',
help='Directory where benchmark is located')
parser.add_argument('--logdir',
dest='logdir',
help='Tensorboard log directory')
parser.add_argument('--ckptdir',
dest='ckptdir',
help='Model checkpoint directory')
# dataset is set by defaults below
# opt and opt_scheduler defined in parse_optimizer
parser.add_argument(
'--max_nodes',
dest='max_nodes',
type=int,
help=
'Maximum number of nodes (ignore graphs with nodes exceeding the number.'
)
parser.add_argument('--cuda', dest='cuda', help='CUDA.')
parser.add_argument('--feature',
dest='feature_type',
help='Feature used for encoder. Can be : id, deg')
# lr and clip defined in parse_optimizer
parser.add_argument('--batch_size',
dest='batch_size',
type=int,
help='Batch size.')
parser.add_argument('--epochs',
dest='num_epochs',
type=int,
help='Number of epochs to train.')
parser.add_argument(
'--train_ratio',
dest='train_ratio',
type=float,
help='Ratio of number of graphs training set to all graphs.')
# test_ratio is set by defaults below
parser.add_argument('--num_workers',
dest='num_workers',
type=int,
help='Number of workers to load data.')
parser.add_argument('--input_dim',
dest='input_dim',
type=int,
help='Input feature dimension')
# Cannot use "parse_prog.py --h" causes "error: ambiguous" (Usage: parse_prog.py --hidden_dim=30)
parser.add_argument('--hidden_dim',
dest='hidden_dim',
type=int,
help='Hidden dimension')
parser.add_argument('--output_dim',
dest='output_dim',
type=int,
help='Output dimension')
parser.add_argument('--num_classes',
dest='num_classes',
type=int,
help='Number of label classes')
parser.add_argument(
'--num_gc_layers',
dest='num_gc_layers',
type=int,
help='Number of graph convolution layers before each pooling')
parser.add_argument('--dropout',
dest='dropout',
type=float,
help='Dropout rate.')
parser.add_argument('--weight_decay',
dest='weight_decay',
type=float,
help='Weight decay regularization constant.')
parser.add_argument('--method',
dest='method',
help='Method. Possible values: base, ')
parser.add_argument('--name-suffix',
dest='name_suffix',
help='suffix added to the output filename')
# assign_ratio is set by defaults below
parser.add_argument('--nobias',
dest='bias',
action='store_const',
const=False,
default=True,
help='Whether to add bias. Default to True.')
parser.add_argument('--gpu',
dest='gpu',
action='store_const',
const=True,
default=False,
help='Whether to use GPU.')
parser.add_argument('--linkpred',
dest='linkpred',
action='store_const',
const=True,
default=False,
help='Whether link prediction side objective is used')
parser.add_argument('--bn',
dest='bn',
action='store_const',
const=True,
default=False,
help='Whether batch normalization is used')
# bmname is set by defaults below
# Set defaults for all program parameters unless provided by user
parser.set_defaults(
datadir="data", # Directory where benchmark is stored (io_parser)
logdir="log", # Tensorboard log directory
ckptdir="ckpt", # Model checkpoint directory
dataset="BAGraph", # Synthetic dataset, syn1
opt="adam", # Opt parser
opt_scheduler="none", # Optimizer scheduler
max_nodes=100, # Maximum number of nodes
# (ignore graphs with nodes exceeding the number)
cuda="0", # CUDA value
feature_type=
"default", # Feature used for encoder with possible values : id, deg
lr=0.001, # Learning rate
clip=2.0, # Gradient clipping
batch_size=20, # Batch size
num_epochs=1000, # Number of epochs to train data
train_ratio=0.8, # Ratio of number of training set to all graphs
test_ratio=0.1,
num_workers=1, # Number of workers to load data
input_dim=10, # Input feature dimension
hidden_dim=20, # Hidden layer dimension
output_dim=20, # Output layer dimension
num_classes=2, # Number of label classes
num_gc_layers=
3, # Number of graph convolution layers before each pooling
dropout=0.0, # Dropout rate
weight_decay=0.005, # Weight decay regularization constant
method="base", # Method used with possible values : base
name_suffix="", # Suffix added to the output filename
assign_ratio=0.1, # Ratio of number of nodes in consecutive layers
bias=True, # "Whether to add bias
gpu=False, # Whether to use GPU
linkpred=False, # Whether link prediction side objective is used
bn=False, # Whether batch normalization is used
bmname=None) # Name of the benchmark datase
return parser.parse_args()
def arg_parse():
parser = argparse.ArgumentParser(description="GNN Explainer arguments.")
io_parser = parser.add_mutually_exclusive_group(required=False)
io_parser.add_argument("--dataset", dest="dataset", help="Input dataset.")
benchmark_parser = io_parser.add_argument_group()
benchmark_parser.add_argument("--bmname",
dest="bmname",
help="Name of the benchmark dataset")
io_parser.add_argument("--pkl",
dest="pkl_fname",
help="Name of the pkl data file")
parser_utils.parse_optimizer(parser)
parser.add_argument(
"--clean-log",
action="store_true",
help="If true, cleans the specified log directory before running.")
parser.add_argument("--logdir",
dest="logdir",
help="Tensorboard log directory")
parser.add_argument("--ckptdir",
dest="ckptdir",
help="Model checkpoint directory")
parser.add_argument("--cuda", dest="cuda", help="CUDA.")
parser.add_argument(
"--gpu",
dest="gpu",
action="store_const",
const=True,
default=False,
help="whether to use GPU.",
)
parser.add_argument("--epochs",
dest="num_epochs",
type=int,
help="Number of epochs to train.")
parser.add_argument("--hidden-dim",
dest="hidden_dim",
type=int,
help="Hidden dimension")
parser.add_argument("--output-dim",
dest="output_dim",
type=int,
help="Output dimension")
parser.add_argument(
"--num-gc-layers",
dest="num_gc_layers",
type=int,
help="Number of graph convolution layers before each pooling",
)
parser.add_argument(
"--bn",
dest="bn",
action="store_const",
const=True,
default=False,
help="Whether batch normalization is used",
)
parser.add_argument("--dropout",
dest="dropout",
type=float,
help="Dropout rate.")
parser.add_argument(
"--nobias",
dest="bias",
action="store_const",
const=False,
default=True,
help="Whether to add bias. Default to True.",
)
parser.add_argument(
"--no-writer",
dest="writer",
action="store_const",
const=False,
default=True,
help="Whether to add bias. Default to True.",
)
# Explainer
parser.add_argument("--mask-act",
dest="mask_act",
type=str,
help="sigmoid, ReLU.")
parser.add_argument(
"--mask-bias",
dest="mask_bias",
action="store_const",
const=True,
default=False,
help="Whether to add bias. Default to True.",
)
parser.add_argument("--explain-node",
dest="explain_node",
type=int,
help="Node to explain.")
parser.add_argument("--graph-idx",
dest="graph_idx",
type=int,
help="Graph to explain.")
parser.add_argument(
"--graph-mode",
dest="graph_mode",
action="store_const",
const=True,
default=False,
help="whether to run Explainer on Graph Classification task.",
)
parser.add_argument(
"--multigraph-class",
dest="multigraph_class",
type=int,
help=
"whether to run Explainer on multiple Graphs from the Classification task for examples in the same class.",
)
parser.add_argument(
"--multinode-class",
dest="multinode_class",
type=int,
help=
"whether to run Explainer on multiple nodes from the Classification task for examples in the same class.",
)
parser.add_argument(
"--align-steps",
dest="align_steps",
type=int,
help="Number of iterations to find P, the alignment matrix.",
)
parser.add_argument("--method",
dest="method",
type=str,
help="Method. Possible values: base, att.")
parser.add_argument("--name-suffix",
dest="name_suffix",
help="suffix added to the output filename")
parser.add_argument(
"--explainer-suffix",
dest="explainer_suffix",
help="suffix added to the explainer log",
)
# TODO: Check argument usage
parser.set_defaults(
logdir="log",
ckptdir="ckpt",
dataset="sat1",
opt="adam",
opt_scheduler="none",
cuda="0",
lr=0.1,
clip=2.0,
batch_size=20,
num_epochs=100,
hidden_dim=20,
output_dim=20,
num_gc_layers=3,
dropout=0.0,
method="base",
name_suffix="",
explainer_suffix="",
align_steps=1000,
explain_node=None,
graph_idx=-1,
mask_act="sigmoid",
multigraph_class=-1,
multinode_class=-1,
)
return parser.parse_args()
def arg_parse():
parser = argparse.ArgumentParser(description='GraphPool arguments.')
io_parser = parser.add_mutually_exclusive_group(required=False)
io_parser.add_argument('--dataset', dest='dataset', help='Input dataset.')
benchmark_parser = io_parser.add_argument_group()
benchmark_parser.add_argument('--bmname',
dest='bmname',
help='Name of the benchmark dataset')
io_parser.add_argument('--pkl',
dest='pkl_fname',
help='Name of the pkl data file')
softpool_parser = parser.add_argument_group()
softpool_parser.add_argument(
'--assign-ratio',
dest='assign_ratio',
type=float,
help='ratio of number of nodes in consecutive layers')
softpool_parser.add_argument('--num-pool',
dest='num_pool',
type=int,
help='number of pooling layers')
parser.add_argument('--linkpred',
dest='linkpred',
action='store_const',
const=True,
default=False,
help='Whether link prediction side objective is used')
parser_utils.parse_optimizer(parser)
parser.add_argument('--datadir',
dest='datadir',
help='Directory where benchmark is located')
parser.add_argument('--logdir',
dest='logdir',
help='Tensorboard log directory')
parser.add_argument('--ckptdir',
dest='ckptdir',
help='Model checkpoint directory')
parser.add_argument('--cuda', dest='cuda', help='CUDA.')
parser.add_argument('--gpu',
dest='gpu',
action='store_const',
const=True,
default=False,
help='whether to use GPU.')
parser.add_argument(
'--max_nodes',
dest='max_nodes',
type=int,
help=
'Maximum number of nodes (ignore graghs with nodes exceeding the number.'
)
parser.add_argument('--batch_size',
dest='batch_size',
type=int,
help='Batch size.')
parser.add_argument('--epochs',
dest='num_epochs',
type=int,
help='Number of epochs to train.')
parser.add_argument(
'--train_ratio',
dest='train_ratio',
type=float,
help='Ratio of number of graphs training set to all graphs.')
parser.add_argument('--num_workers',
dest='num_workers',
type=int,
help='Number of workers to load data.')
parser.add_argument('--feature',
dest='feature_type',
help='Feature used for encoder. Can be: id, deg')
parser.add_argument('--input_dim',
dest='input_dim',
type=int,
help='Input feature dimension')
parser.add_argument('--hidden_dim',
dest='hidden_dim',
type=int,
help='Hidden dimension')
parser.add_argument('--output_dim',
dest='output_dim',
type=int,
help='Output dimension')
parser.add_argument('--num_classes',
dest='num_classes',
type=int,
help='Number of label classes')
parser.add_argument(
'--num_gc_layers',
dest='num_gc_layers',
type=int,
help='Number of graph convolution layers before each pooling')
parser.add_argument('--bn',
dest='bn',
action='store_const',
const=True,
default=False,
help='Whether batch normalization is used')
parser.add_argument('--dropout',
dest='dropout',
type=float,
help='Dropout rate.')
parser.add_argument('--nobias',
dest='bias',
action='store_const',
const=False,
default=True,
help='Whether to add bias. Default to True.')
parser.add_argument('--weight_decay',
dest='weight_decay',
type=float,
help='Weight decay regularization constant.')
parser.add_argument('--method',
dest='method',
help='Method. Possible values: base, ')
parser.add_argument('--name-suffix',
dest='name_suffix',
help='suffix added to the output filename')
parser.set_defaults(
datadir='data', # io_parser
logdir='log',
ckptdir='ckpt',
dataset='syn1',
opt='adam', # opt_parser
opt_scheduler='none',
max_nodes=100,
cuda='1',
feature_type='default',
lr=0.0005,
clip=2.0,
batch_size=20,
num_epochs=1000,
train_ratio=0.8,
test_ratio=0.1,
num_workers=1,
input_dim=342,
hidden_dim=20,
output_dim=20,
num_classes=4,
num_gc_layers=10,
dropout=0.0,
weight_decay=0.005,
method='base',
name_suffix='',
assign_ratio=0.1,
)
return parser.parse_args()
def arg_parse():
print("Attempt to parse arguments for GNN Explainer ...")
parser = argparse.ArgumentParser(description='GNN program arguments.')
# Set optimizer parameters
parser_utils.parse_optimizer(parser)
## Add parsing arguments
parser.add_argument('--logdir', dest='logdir',
help='Tensorboard log directory')
parser.add_argument('--ckptdir', dest='ckptdir',
help='Model checkpoint directory')
# dataset is set by defaults below
# opt and opt_scheduler defined in parse_optimizer
parser.add_argument("--cuda", dest="cuda", help="CUDA.")
# lr and clip defined in parse_optimizer
parser.add_argument('--batch_size', dest='batch_size', type=int,
help='Batch size.')
parser.add_argument('--epochs', dest='num_epochs', type=int,
help='Number of epochs to train.')
# Cannot use "parse_prog.py --h" causes "error: ambiguous" (Usage: parse_prog.py --hidden_dim=30)
parser.add_argument('--hidden_dim', dest='hidden_dim', type=int,
help='Hidden dimension')
parser.add_argument('--output_dim', dest='output_dim', type=int,
help='Output dimension')
parser.add_argument('--num_classes', dest='num_classes', type=int,
help='Number of label classes')
parser.add_argument('--num_gc_layers', dest='num_gc_layers', type=int,
help='Number of graph convolution layers before each pooling')
parser.add_argument('--dropout', dest='dropout', type=float,
help='Dropout rate.')
parser.add_argument('--method', dest='method',
help='Method. Possible values: base, att.')
parser.add_argument('--name-suffix', dest='name_suffix',
help='suffix added to the output filename')
parser.add_argument('--nobias', dest='bias', action='store_const',
const=False, default=True,
help='Whether to add bias. Default to True.')
parser.add_argument('--gpu', dest='gpu', action='store_const',
const=True, default=False, help='Whether to use GPU.')
parser.add_argument('--bn', dest='bn', action='store_const',
const=True, default=False,
help='Whether batch normalization is used')
# bmname is set by defaults below
parser.add_argument("--no-writer", dest="writer", action="store_const",
const=False, default=True,
help="Whether to add bias. Default to True.")
# Parameters for Explainer
parser.add_argument("--mask-act", dest="mask_act", type=str,
help="sigmoid, ReLU.")
parser.add_argument("--mask-bias", dest="mask_bias", action="store_const",
const=True, default=False,
help="Whether to add bias. Default to True.")
parser.add_argument("--explain-node", dest="explain_node", type=int,
help="Node to explain.")
parser.add_argument("--graph-idx", dest="graph_idx", type=int,
help="Graph to explain.")
parser.add_argument("--graph-mode", dest="graph_mode", action="store_const",
const=True, default=False,
help="Whether to run Explainer on Graph Classification task.")
parser.add_argument("--multigraph-class", dest="multigraph_class", type=int,
help="Whether to run Explainer on multiple Graphs from the Classification task for examples in the same class.")
parser.add_argument("--multinode-class", dest="multinode_class", type=int,
help="whether to run Explainer on multiple nodes from the Classification task for examples in the same class.")
parser.add_argument("--explainer-suffix", dest="explainer_suffix",
help="suffix added to the explainer log")
# Set defaults for all program parameters unless provided by user
parser.set_defaults(logdir = "log", # Tensorboard log directory
ckptdir = "ckpt", # Model checkpoint directory
dataset = "BAGraph", # Synthetic dataset, syn1
opt = "adam", # Opt parser
opt_scheduler = "none", # Optimizer scheduler
cuda = "0", # CUDA value
lr = 0.1, # Learning rate (train : 0.001)
clip = 2.0, # Gradient clipping
batch_size = 20, # Batch size
num_epochs = 100, # Number of epochs for explainer training (train : 1000)
hidden_dim = 20, # Hidden layer dimension
output_dim = 20, # Output layer dimension
num_classes = 2, # Number of label classes
num_gc_layers = 3, # Number of graph convolution layers before each pooling
dropout = 0.0, # Dropout rate
method = "base", # Method used with possible values : base
name_suffix = "", # Suffix added to the output filename
bias = True, # "Whether to add bias
gpu = False, # Whether to use GPU
bn = False, # Whether batch normalization is used
bmname = None, # Name of the benchmark datase
explainer_suffix="", # Suffix added to the explainer log
explain_node = 25, # Node to explain (eg. Check is on None)
graph_idx = -1, # Graph to explain
mask_act="sigmoid", # Type of activation for mask (sigmoid, ReLU)
multigraph_class = -1, # Whether to run Explainer on multiple Graphs from the Classification task for examples in the same class
multinode_class=-1) # Whether to run Explainer on multiple nodes from the Classification task for examples in the same class
return parser.parse_args()