optimizer = torch.optim.Adam(net.parameters(),
eps=0.000001,
lr=args.lr,
betas=(0.9, 0.999),
weight_decay=0.0001)
# training loop
for epoch in range(1, args.epochs):
accuracy = train(net, train_loader, optimizer)
if epoch % 100 == 0 and accuracy == 100:
break
# graph hidden units
for layer in [1, 2]:
if layer == 1 or args.net != 'polar':
for node in range(args.hid):
graph_hidden(net, layer, node)
plt.scatter(full_input[:, 0],
full_input[:, 1],
c=1 - full_target[:, 0],
cmap='RdYlBu')
plt.savefig('%s%d_%d.png' % (args.net, layer, node))
# graph output unit
graph_output(net)
plt.scatter(full_input[:, 0],
full_input[:, 1],
c=1 - full_target[:, 0],
cmap='RdYlBu')
plt.savefig('%s_out.png' % args.net)
optimizer = torch.optim.Adam(net.parameters(),
eps=0.000001,
lr=args.lr,
betas=(0.9, 0.999),
weight_decay=0.0001)
# optimizer = torch.optim.SGD(net.parameters(), lr=args.lr, weight_decay=0.0001)
for epoch in range(1, args.epochs):
accuracy = train(net, train_loader, optimizer)
if epoch % 100 == 0 and accuracy == 100:
break
# save model
for layer in [1, 2]:
if layer == 1 or args.net != 'polar':
for node in range(args.hid):
graph_hidden(net, layer, node) # output hiden layer values
plt.scatter(full_input[:, 0],
full_input[:, 1],
c=1 - full_target[:, 0],
cmap='RdYlBu')
plt.savefig('%s%d_%d.png' % (args.net, layer, node))
graph_output(net)
plt.scatter(full_input[:, 0],
full_input[:, 1],
c=1 - full_target[:, 0],
cmap='RdYlBu')
plt.savefig('%s_out.png' % args.net)