in python/dgllife/model/pretrain/moleculenet/toxcast.py [0:0]
def create_toxcast_model(model_name):
"""Create a model.
Parameters
----------
model_name : str
Name for the model.
Returns
-------
Created model
"""
n_tasks = 617
if model_name == 'GCN_canonical_ToxCast':
num_gnn_layers = 3
dropout = 0.2354934715188964
return GCNPredictor(in_feats=74,
hidden_feats=[256] * num_gnn_layers,
activation=[F.relu] * num_gnn_layers,
residual=[True] * num_gnn_layers,
batchnorm=[False] * num_gnn_layers,
dropout=[dropout] * num_gnn_layers,
predictor_hidden_feats=512,
predictor_dropout=dropout,
n_tasks=n_tasks)
elif model_name == 'GCN_attentivefp_ToxCast':
dropout = 0.5790202973197223
return GCNPredictor(in_feats=39,
hidden_feats=[256],
activation=[F.relu],
residual=[False],
batchnorm=[True],
dropout=[dropout],
predictor_hidden_feats=16,
predictor_dropout=dropout,
n_tasks=n_tasks)
elif model_name == 'GAT_canonical_ToxCast':
num_gnn_layers = 3
dropout = 0.30477898651808644
return GATPredictor(in_feats=74,
hidden_feats=[256] * num_gnn_layers,
num_heads=[6] * num_gnn_layers,
feat_drops=[dropout] * num_gnn_layers,
attn_drops=[dropout] * num_gnn_layers,
alphas=[0.5850073967467644] * num_gnn_layers,
residuals=[True] * num_gnn_layers,
biases=[False] * num_gnn_layers,
predictor_hidden_feats=256,
predictor_dropout=dropout,
n_tasks=n_tasks)
elif model_name == 'GAT_attentivefp_ToxCast':
num_gnn_layers = 3
dropout = 0.039304570924327153
return GATPredictor(in_feats=39,
hidden_feats=[32] * num_gnn_layers,
num_heads=[8] * num_gnn_layers,
feat_drops=[dropout] * num_gnn_layers,
attn_drops=[dropout] * num_gnn_layers,
alphas=[0.8044239663965763] * num_gnn_layers,
residuals=[False] * num_gnn_layers,
biases=[False] * num_gnn_layers,
predictor_hidden_feats=128,
predictor_dropout=dropout,
n_tasks=n_tasks)
elif model_name == 'Weave_canonical_ToxCast':
return WeavePredictor(node_in_feats=74,
edge_in_feats=13,
num_gnn_layers=1,
gnn_hidden_feats=256,
graph_feats=256,
gaussian_expand=False,
n_tasks=n_tasks)
elif model_name == 'Weave_attentivefp_ToxCast':
return WeavePredictor(node_in_feats=39,
edge_in_feats=11,
num_gnn_layers=1,
gnn_hidden_feats=128,
graph_feats=128,
gaussian_expand=False,
n_tasks=n_tasks)
elif model_name == 'MPNN_canonical_ToxCast':
return MPNNPredictor(node_in_feats=74,
edge_in_feats=13,
node_out_feats=32,
edge_hidden_feats=32,
num_step_message_passing=2,
num_step_set2set=2,
num_layer_set2set=3,
n_tasks=n_tasks)
elif model_name == 'MPNN_attentivefp_ToxCast':
return MPNNPredictor(node_in_feats=39,
edge_in_feats=11,
node_out_feats=32,
edge_hidden_feats=64,
num_step_message_passing=5,
num_step_set2set=3,
num_layer_set2set=3,
n_tasks=n_tasks)
elif model_name == 'AttentiveFP_canonical_ToxCast':
return AttentiveFPPredictor(node_feat_size=74,
edge_feat_size=13,
num_layers=1,
num_timesteps=1,
graph_feat_size=64,
dropout=0.00020996203951184622,
n_tasks=n_tasks)
elif model_name == 'AttentiveFP_attentivefp_ToxCast':
return AttentiveFPPredictor(node_feat_size=39,
edge_feat_size=11,
num_layers=1,
num_timesteps=3,
graph_feat_size=32,
dropout=0.15914067005489962,
n_tasks=n_tasks)
elif model_name == 'gin_supervised_contextpred_ToxCast':
jk = 'last'
model = GINPredictor(
num_node_emb_list=[120, 3],
num_edge_emb_list=[6, 3],
num_layers=5,
emb_dim=300,
JK=jk,
dropout=0.5,
readout='max',
n_tasks=n_tasks
)
model.gnn.JK = jk
return model
elif model_name == 'gin_supervised_infomax_ToxCast':
jk = 'sum'
model = GINPredictor(
num_node_emb_list=[120, 3],
num_edge_emb_list=[6, 3],
num_layers=5,
emb_dim=300,
JK=jk,
dropout=0.5,
readout='sum',
n_tasks=n_tasks
)
model.gnn.JK = jk
return model
elif model_name == 'gin_supervised_edgepred_ToxCast':
jk = 'sum'
model = GINPredictor(
num_node_emb_list=[120, 3],
num_edge_emb_list=[6, 3],
num_layers=5,
emb_dim=300,
JK=jk,
dropout=0.5,
readout='attention',
n_tasks=n_tasks
)
model.gnn.JK = jk
return model
elif model_name == 'gin_supervised_masking_ToxCast':
jk = 'sum'
model = GINPredictor(
num_node_emb_list=[120, 3],
num_edge_emb_list=[6, 3],
num_layers=5,
emb_dim=300,
JK=jk,
dropout=0.5,
readout='max',
n_tasks=n_tasks
)
model.gnn.JK = jk
return model
elif model_name == 'NF_canonical_ToxCast':
num_gnn_layers = 4
dropout = 0.23659695490588817
return NFPredictor(in_feats=74,
n_tasks=n_tasks,
hidden_feats=[128] * num_gnn_layers,
batchnorm=[True] * num_gnn_layers,
dropout=[dropout] * num_gnn_layers,
predictor_hidden_size=512,
predictor_batchnorm=True,
predictor_dropout=dropout)
else:
return None