desc += 'acc: %.3f, TP: %.3f, TN: %.3f, FN: %.3f, FP: %.3f' % (
(TP.item() + TN.item()) * 1.0 / TOT.item(), TP.item() * 1.0 / TOT.item(),
TN.item() * 1.0 / TOT.item(), FN.item() * 1.0 / TOT.item(),
FP.item() * 1.0 / TOT.item())
running_loss += loss.item() * n_batches
nbre_sample += n_batches
#print(ok)
epoch_loss = running_loss / nbre_sample
acc = (TP.item() + TN.item()) * 1.0 / TOT.item()
nn_model_ref.acc = acc
print('{} Loss: {:.4f}'.format(
phase, epoch_loss))
print('{} Acc: {:.4f}'.format(
phase, acc))
#print(desc)
print()
time_elapsed = time.time() - since
print('Evaluation complete in {:.0f}m {:.0f}s'.format(
time_elapsed // 60, time_elapsed % 60))
print()
from sklearn.model_selection import cross_val_score
from sklearn import svm
#del data_ici
#nn_model_ref2.intermediaires[phase].append(nn_model_ref2.net.intermediare.detach().cpu().numpy().astype(np.uint8))
#nn_model_ref2.outputs_proba[phase].append(outputs.detach().cpu().numpy().astype(np.float16))
nbre_sample += n_batches
epoch_loss = running_loss / nbre_sample
acc1 = (correct1.item()) * 1.0 / TOT21.item()
acc2 = (correct2.item()) * 1.0 / TOT22.item()
acc3 = (correct3.item()) * 1.0 / TOT23.item()
acc4 = (correct4.item()) * 1.0 / TOT24.item()
#acc = (correct1.item() + correct2.item()) * 1.0 / (TOT22.item() + TOT21.item())
acc = (correct1.item() + correct2.item() + correct3.item() +
correct4.item()) * 1.0 / (TOT22.item() + TOT21.item() +
TOT23.item() + TOT24.item())
nn_model_ref2.acc = acc
print('{} Loss: {:.4f}'.format(phase, epoch_loss))
print('{} Acc: {:.4f}'.format(phase, acc1))
print('{} Acc: {:.4f}'.format(phase, acc2))
print('{} Acc: {:.4f}'.format(phase, acc3))
print('{} Acc: {:.4f}'.format(phase, acc4))
print('{} Acc FINAL: {:.4f}'.format(phase, acc))
#print(desc)
print()
time_elapsed = time.time() - since
print('Evaluation complete in {:.0f}m {:.0f}s'.format(
time_elapsed // 60, time_elapsed % 60))
print()
num1 = int(nn_model_ref2.args.nbre_sample_train_classifier /
nn_model_ref2.batch_size)
num2 = int(nn_model_ref2.args.nbre_sample_val_classifier /