def save_results_dem(self):
# plt.plot(np.arange(1,len(self.time_complex)+1),self.time_complex)
# plt.show()
root_train = np.asarray(self.gs_level_train)[K_Levels, :, 0]
root_test = np.asarray(self.gs_level_test)[K_Levels, :, 0]
# write_file(RS_PATH+self.args.algorithm+"_"+ complex_file_path, time_complex=self.time_complex)
# print('Saving complexity..\nOK!')
rs_path = RS_PATH + self.args.algorithm + "_" + rs_file_path
write_file(
file_name=rs_path,
root_test=root_test,
root_train=root_train,
cs_avg_data_test=self.cs_avg_data_test,
cs_avg_data_train=self.cs_avg_data_train,
cg_avg_data_test=self.cg_avg_data_test,
cg_avg_data_train=self.cg_avg_data_train,
cs_data_test=self.cs_data_test,
cs_data_train=self.cs_data_train,
cg_data_test=self.cg_data_test,
cg_data_train=self.cg_data_train,
gs_level_train=self.gs_level_train,
gs_level_test=self.gs_level_test,
gg_level_train=self.gg_level_train,
gg_level_test=self.gg_level_test,
gks_level_train=self.gks_level_train,
gks_level_test=self.gks_level_test,
gkg_level_train=self.gkg_level_train,
gkg_level_test=self.gkg_level_test,
dendo_data=self.dendo_data,
dendo_data_round=self.dendo_data_round, # Dendrogram data
N_clients=[N_clients],
TREE_UPDATE_PERIOD=[TREE_UPDATE_PERIOD]) # Setting
plot_from_file(rs_path)
def train_model(model, trainloader):
learning_rate = 0.001
num_epochs = 40
epochs_stop = 5
min_loss = np.Inf
epoch_min_loss = np.Inf
no_improve = 0
criterion = torch.nn.CrossEntropyLoss()
optimizer = Adam(model.parameters(), lr=learning_rate)
total_step = len(trainloader)
loss_list = []
acc_list = []
for epoch in range(num_epochs):
for i, (labels, sounds) in enumerate(trainloader):
if torch.cuda.is_available():
labels = labels.cuda()
sounds = sounds.cuda()
# Run the forward pass
outputs = model(sounds)
loss = criterion(outputs, labels)
loss_list.append(loss.item())
if loss < min_loss:
min_loss = loss
# Backprop and perform Adam optimization
optimizer.zero_grad()
loss.backward()
optimizer.step()
# Track the accuracy
total = labels.size(0)
_, predicted = torch.max(outputs.data, 1)
correct = (predicted == labels).sum().item()
acc_list.append(correct / total)
if (i + 1) % 10 in [0, 10]:
print(
'Epoch [{}/{}], Step [{}/{}], Loss: {:.4f}, Accuracy: {:.2f}%'
.format(epoch + 1, num_epochs, i + 1, total_step,
loss.item(), (correct / total) * 100))
if min_loss < epoch_min_loss:
epoch_min_loss = min_loss
no_improve = 0
else:
no_improve += 1
if no_improve == epochs_stop:
break
else:
continue
loss_list.append('-----' + str(epoch) + '-----')
write_file('accuracy.txt', acc_list)
write_file('loss.txt', loss_list)
def save_results1(self):
rs_path = RS_PATH + rs_file_path
write_file(file_name=rs_path,
root_test=self.rs_glob_acc,
root_train=self.rs_train_acc,
cs_avg_data_test=self.cs_avg_data_test,
cs_avg_data_train=self.cs_avg_data_train,
cg_avg_data_test=self.cg_avg_data_test,
cg_avg_data_train=self.cg_avg_data_train,
cs_data_test=self.cs_data_test,
cs_data_train=self.cs_data_train,
cg_data_test=self.cg_data_test,
cg_data_train=self.cg_data_train,
N_clients=[N_clients])
plot_from_file(rs_path)
def test_model(model, testloader):
correct = 0
total = 0
y = []
y_predicted = []
with torch.no_grad():
for labels, sounds in testloader:
if torch.cuda.is_available():
labels = labels.cuda()
sounds = sounds.cuda()
outputs = model(sounds)
_, predicted = torch.max(outputs.data, 1)
total += labels.size(0)
correct += (predicted == labels).sum().item()
y_predicted.append(outputs.numpy())
y.append(labels.numpy())
print('Accuracy: %d %%' % (100 * correct / total))
write_file('accuracy_test.txt', [(100 * correct / total)])
torch.save(model.state_dict(), '.')