def animate(i):
#for epoch in range(1,no_epochs+1):
epoch = i + 1
if epoch < 10:
epoch_str = '000' + str(epoch)
elif epoch >= 10 and epoch < 100:
epoch_str = '00' + str(epoch)
elif epoch >= 100 and epoch < 1000:
epoch_str = '0' + str(epoch)
else:
epoch_str = str(epoch)
model_path = model_path_for_video + '_' + epoch_str + '.pt'
print('Loading...' + model_path)
sig_model = network_models.load_saved_model(
model_path,
model_type,
input_size,
kernel_size,
filter_number=filter_number,
signal_number=len(sig_type_source),
no_layers=no_layers)
Y_batch_predicted = sig_model.forward(
X_batch).squeeze().detach().cpu().numpy()
#for v in range(Y_batch.shape[0]):
plt.clf()
plt.subplot(no_sigs + 1, 1, 1)
#plt.plot(Y_batch[v,0,:] , '-r')
plt.plot((Y_batch[v, 0, :] - np.mean(Y_batch[v, 0, :])) / (np.sqrt(
np.sum(np.power(Y_batch[v, 0, :] - np.mean(Y_batch[v, 0, :]),
2)))), '-r')
#plt.plot((Y_batch_predicted[v,:]-np.min(Y_batch_predicted[v,:]) )/(np.max(Y_batch_predicted[v,:])-np.min(Y_batch_predicted[v,:])) , '-b')
plt.plot((Y_batch_predicted[v, :] - np.mean(Y_batch_predicted[v, :])) /
(np.sqrt(
np.sum(
np.power(
Y_batch_predicted[v, :] -
np.mean(Y_batch_predicted[v, :]), 2)))), '-b')
#plt.plot(Y_batch_predicted[v, :] , '-b')
# plt.title('Subject Number: ' + str(list_subjects[v]) +' , Epoch: ' + str(epoch)
# + ', Train Loss=' + str(train_history[epoch-1]['train_loss']) + ', Val Loss=' + str(valid_history[epoch-1]['valid_loss']) )
plt.title(
'Subject Number: {}, Epoch: {}, Train Loss: {:.4f}, Val Loss: {:.4f}'
.format(list_subjects[v], epoch,
train_history[epoch - 1]['train_loss'],
valid_history[epoch - 1]['valid_loss']))
for k in range(2, no_sigs + 2):
plt.subplot(no_sigs + 1, 1, k)
plt.plot(X_batch_detached[v, k - 2, :], '-g')
plt.plot(
(Y_batch_predicted[v, :] - np.min(Y_batch_predicted[v, :])) /
(np.max(Y_batch_predicted[v, :]) -
np.min(Y_batch_predicted[v, :])), '-b')
#plt.plot(Y_batch_predicted[v,:] , '-b')
plt.title(sig_type_source[k - 2])
# plt.tight_layout()
# plt.show()
# if save_true:
# fig.savefig(directory + '/' + file_name_pre + '_waveforms_' + str(v)+ '.png')
del Y_batch_predicted
input_size = frame_length_testing//down_sample_factor_testing
#create the testing generator
test_gen=data_generators.make_generator_multiple_signal_testing(list_of_subjects=all_testing_subject_instances, batch_size=batch_size, eps=eps, frame_length=frame_length_testing
,stride=stride_testing, list_sig_type_source=sig_type_source+['ecg'], sig_type_target=sig_type_target
, down_sample_factor=down_sample_factor_testing, normalized=normalized, store_in_ram=True)
#check generator !!
#data_generators.diagnose_generator_multiple_signal_testing(test_gen , sig_type_source+['ecg'])
#load model
model_path = directory + '/Code Output/'+ file_name_pre + '.pt'
model_type=config['model_type']
kernel_size=config['kernel_size']
filter_number=config['filter_number']
sig_model= network_models.load_saved_model(model_path , model_type , input_size , kernel_size , filter_number=filter_number, signal_number=len(sig_type_source),
no_layers=no_layers)
def test_model(gen, sig_model):
'''
function tests a model by running it over the test data and calculating error metrics
:param gen: test generator
:param sig_model: model
:return: list_pearson_r_loss: list of pearson correlations between target and estimated segments
:return: list_subject_ids: subject ID for each segment
:return: list_i_errors: list of R-I interval errors between each target and estimated signal segment
:return: list_j_errors: list of R-J interval errors between each target and estimated signal segment
:return: list_k_errors: list of R-K interval errors between each target and estimated signal segment
'''
criterion = network_models.PearsonRLoss()
with torch.no_grad():
def main():
#sampling rate of training dataset
F_SAMPLING=2000
#config
parser = argparse.ArgumentParser()
arg = parser.add_argument
arg('--n_f', type=int, help='n_f')
arg('--L', type=int , help='L')
args = vars(parser.parse_args())
print(args)
#configs
config = {
'file_name_pre': 'mdl_00000_a000_dd_mmm_nn_'+str(args['L'])+'_'+str(args['n_f']),
'no_layers': args['L'] ,
'filter_number': args['n_f'],
'sig_type_source': ['aX', 'aY', 'aZ'],
'mode':'both',
'eps': 1e-3,
'model_type': 'Unetxl',
'down_sample_factor':4,
'frame_length' : int(4.096*F_SAMPLING),
'kernel_size': 7 , #(3,5) , #5,#(3,5), #5, #(3, 5),#5
'directory':'/media/sinan/9E82D1BB82D197DB/RESEARCH VLAB work on/Gyroscope SCG project/Deep Learning Paper Code and Materials',
'cycle_per_batch':2,
'sig_type_target': 'bcg',
'loss_func':'pearson_r',
'produce_video' : False,
'store_in_ram':True,
'augment_accel':True,
'augment_theta_lim': 10,
'augment_prob':0.5
}
#
cycle_per_batch = config['cycle_per_batch']
mode= config['mode']
eps = config['eps']
kernel_size = config['kernel_size']
directory= config['directory']
model_type = config['model_type']
no_layers=config['no_layers']
filter_number=config['filter_number']
sig_type_source=config['sig_type_source']
sig_type_target=config['sig_type_target']
down_sample_factor = config['down_sample_factor']
frame_length=config['frame_length']
input_size = frame_length//down_sample_factor
normalized = True if model_type!='Unet_multiple_signal_in_not_normalized' else False
loss_func = config['loss_func']
produce_video= config['produce_video']
store_in_ram=config['store_in_ram']
augment_accel= config['augment_accel']
augment_theta_lim = config['augment_theta_lim']
augment_prob=config['augment_prob']
file_name_pre = config['file_name_pre']
file_name_pre = file_name_pre[0:4] + model_type[:-1] + str(no_layers) +file_name_pre[9::]
axis_string = ''.join(['x' if 'aX' in sig_type_source else '0' , 'y' if 'aY' in sig_type_source else '0' , 'z' if 'aZ' in sig_type_source else '0' ])
file_name_pre = file_name_pre[:12] + axis_string + file_name_pre[15::]
print('Model Name: ' + file_name_pre)
#get all subject data
all_subject_instances = utility.load_subjects(directory + '/Training Data Analog Acc', store_in_ram )
#train test split
train_subject_instances, val_subject_instances = train_test_split( all_subject_instances , test_size=0.2, random_state=49 )
#make a train and val generator
train_gen = data_generators.make_generator_multiple_signal(list_of_subjects=train_subject_instances, cycle_per_batch=cycle_per_batch, eps=eps,frame_length=frame_length,
mode=mode, list_sig_type_source= sig_type_source, sig_type_target= sig_type_target , down_sample_factor =down_sample_factor,
normalized=normalized , store_in_ram=store_in_ram ,
augment_accel = augment_accel , augment_theta_lim = augment_theta_lim , augment_prob=augment_prob)
val_gen = data_generators.make_generator_multiple_signal(list_of_subjects=val_subject_instances, cycle_per_batch=cycle_per_batch, eps=eps, frame_length=frame_length,
mode=mode, list_sig_type_source= sig_type_source, sig_type_target= sig_type_target, down_sample_factor=down_sample_factor,
normalized=normalized, store_in_ram=store_in_ram)
#check !
#utility.diagnose_training_subjects(all_subject_instances)
#data_generators.diagnose_generator_multiple_signal(train_gen , sig_type_source)
#data_generators.diagnose_generator_multiple_signal(val_gen , sig_type_source)
#create model
if model_type=='Unetxl':
sig_model = network_models.Unet_xl(input_size, kernel_size, filter_number, len(sig_type_source) , no_layers )
#make model parallel
sig_model = nn.DataParallel(sig_model.cuda(), device_ids=[0,1])
#loss function is negative pearson loss
loss = network_models.PearsonRLoss()
#training configs
cudnn.benchmark = True
config['n_epochs'] = 150
config['scheduler_milestones'] = [30,60,120] #[50,100,200]
config['train_steps'] = 10
config['val_steps'] = 10
config['initial_lr'] = 0.001
config['model_path'] = directory + '/Code Output/best_so_far.pt'
config['model_path_for_video'] = directory + '/Models for Video/' + file_name_pre
args = {'lr': config['initial_lr'],
'n_epochs':config['n_epochs'],
'model_path': config['model_path'],
'step_count':config['train_steps'],
'val_steps': config['val_steps'],
'scheduler_milestones': config['scheduler_milestones'],
'model_path_for_video': config['model_path_for_video']
}
#train the model
start_model_train = time.time()
train_history, valid_history , best_val= network_models.train_torch_generator_with_video(args=args,
sig_model=sig_model,
criterion=loss,
train_gen=train_gen,
val_gen=val_gen,
init_optimizer = lambda lr: Adam(sig_model.parameters(), lr=lr),
init_schedule = lambda optimizer,milestones: torch.optim.lr_scheduler.MultiStepLR(optimizer, milestones=milestones, gamma=0.5) ,
produce_video=produce_video)
end_model_train = time.time()
print('Model Training Duration In Seconds: ' + str(end_model_train - start_model_train))
print('Best Validation Loss: ' + str(best_val))
#save model
sig_model = network_models.load_saved_model(model_path=config['model_path'],
model_type= model_type,
input_size=input_size ,
kernel_size=kernel_size ,
filter_number=filter_number,
signal_number=len(sig_type_source),
no_layers = no_layers)
torch.save({
'model': sig_model.state_dict(),
}, directory+'/Code Output/' + file_name_pre + '.pt')
#save workspace
pickle_list = [config, train_history , valid_history, train_subject_instances, val_subject_instances , best_val]
fileObject = open(directory+'/Code Output/' +file_name_pre+'_pickle','wb')
pickle.dump(pickle_list,fileObject)
fileObject.close()
#print configs to a test file
with open(directory + '/Code Output/' + file_name_pre + '_config.txt', "w") as text_file:
print(config, file=text_file)
print('Model Training Duration In Seconds: ' + str(end_model_train - start_model_train), file=text_file)
#plot results
network_models.show_loss_torch_model(train_history, valid_history, file_name_pre , directory+'/Code Output')
