def get_model(cfg):
num_classes = int(cfg['n_classes'])
input_size = int(cfg['data_dim'])
if cfg['model'] == 'blstm':
classifier = BiLSTM(input_size,
n_classes=num_classes,
embedding_size=128,
hidden_size=256,
dropout=cfg['dropout'])
if cfg['model'] == 'sdec':
classifier = DECSeq(
input_size,
int(cfg['embedding_size']),
num_classes,
#dropout=cfg['dropout'],
k=5,
aggr='max',
pool_op=cfg['pool_op'])
if cfg['model'] == 'dec':
classifier = DEC(
input_size,
int(cfg['embedding_size']),
num_classes,
#dropout=cfg['dropout'],
k=5,
aggr='max',
pool_op=cfg['pool_op'])
elif cfg['model'] == 'pn_geom':
classifier = PNptg2(input_size,
int(cfg['embedding_size']),
num_classes,
same_size=cfg['same_size'])
return classifier
def get_model(cfg):
num_classes = int(cfg['n_classes'])
input_size = int(cfg['data_dim'])
if cfg['model'] == 'blstm':
classifier = BiLSTM(input_size,
n_classes=num_classes,
embedding_size=128,
hidden_size=256,
dropout=cfg['dropout'])
if cfg['model'] == 'dec_ori':
classifier = DGCNNSeq(input_size,
int(cfg['embedding_size']),
num_classes,
batch_size=int(cfg['batch_size']),
k=5,
fov=1,
dropout=0.5)
if cfg['model'] == 'dec':
classifier = DECSeq(
input_size,
int(cfg['embedding_size']),
num_classes,
dropout=cfg['dropout'],
#fov=3,
k=int(cfg['k_dec']),
# k=5,
aggr='max',
#pool_op=global_max_pool)
pool_op=cfg['pool_op'])
#bn=True)
if cfg['model'] == 'nnc':
classifier = NNC(input_size,
int(cfg['embedding_size']),
num_classes,
batch_size=int(cfg['batch_size']),
pool_op=cfg['pool_op'],
same_size=cfg['same_size'])
if cfg['model'] == 'gcn':
classifier = GCNConvNetBN(input_size,
int(cfg['embedding_size']),
num_classes,
pool_op=cfg['pool_op'],
same_size=cfg['same_size'])
if cfg['model'] == 'gat':
classifier = GAT(input_size,
num_classes)
elif cfg['model'] == 'pn_geom':
classifier = PNptg2(input_size,
int(cfg['embedding_size']),
num_classes,
#batch_size=int(cfg['batch_size']),
same_size=cfg['same_size'])
return classifier
def get_model(cfg):
num_classes = int(cfg['n_classes'])
input_size = int(cfg['data_dim'])
n_gf = int(cfg['num_gf'])
if cfg['model'] == 'blstm':
classifier = BiLSTM(input_size,
n_classes=num_classes,
embedding_size=128,
hidden_size=256)
if cfg['model'] == 'dec_ori':
classifier = DGCNNSeq(input_size,
int(cfg['embedding_size']),
num_classes,
batch_size=int(cfg['batch_size']),
k=3,
fov=1,
dropout=0.5)
if cfg['model'] == 'dec':
classifier = DECSeq(
input_size,
int(cfg['embedding_size']),
num_classes,
#fov=3,
batch_size=int(cfg['batch_size']),
k=7,
aggr='max',
pool_op=global_max_pool,
same_size=cfg['same_size'])
#bn=True)
if cfg['model'] == 'nnc':
classifier = NNC(input_size,
int(cfg['embedding_size']),
num_classes,
batch_size=int(cfg['batch_size']),
pool_op=global_max_pool,
same_size=cfg['same_size'])
if cfg['model'] == 'gcn':
classifier = GCNConvNet(input_size,
int(cfg['embedding_size']),
num_classes,
batch_size=int(cfg['batch_size']),
pool_op=global_max_pool,
same_size=cfg['same_size'])
elif cfg['model'] == 'pn_geom':
classifier = PNptg(input_size,
int(cfg['embedding_size']),
num_classes,
batch_size=int(cfg['batch_size']),
pool_op=global_max_pool,
same_size=cfg['same_size'])
return classifier