current_means = class_means[:, order[:(iteration+1)*nb_cl]]
causal_embed = np.load(ckp_path+'/causal_embed_run_{}_iteration_{}.npy'.format(args.run_id, iteration))
causal_embed = torch.from_numpy(causal_embed).to(device)
causal_embed = causal_embed/(torch.norm(causal_embed, 2, 1, keepdim=True))
# each task
for task_i in range(start_iter, iteration+1):
if task_i == start_iter:
indices = np.array([i in range(0, (task_i+1)*nb_cl) for i in map_input_labels])
else:
indices = np.array([i in range(task_i*nb_cl, (task_i+1)*nb_cl) for i in map_input_labels])
current_eval_set = merge_images_labels(input_data[indices], map_input_labels[indices])
evalset.imgs = evalset.samples = current_eval_set
evalloader = torch.utils.data.DataLoader(evalset, batch_size=128,
shuffle=False, num_workers=8, pin_memory=True)
acc = compute_accuracy_TDE(tg_model, tg_feature_model, current_means,
evalloader, causal_embed, iteration=iteration-start_iter, print_info=False)
all_accu_matrix[step_i, task_i-start_iter] = acc
# cumul
indices = np.array([i in range(0, (iteration+1)*nb_cl) for i in map_input_labels])
current_eval_set = merge_images_labels(input_data[indices], map_input_labels[indices])
evalset.imgs = evalset.samples = current_eval_set
evalloader = torch.utils.data.DataLoader(evalset, batch_size=128,
shuffle=False, num_workers=8, pin_memory=True)
acc = compute_accuracy_TDE(tg_model, tg_feature_model, current_means,
evalloader, causal_embed, iteration=iteration-start_iter, print_info=False)
all_accu_matrix[step_i, all_step_num] = acc
print(acc)
step_i += 1
batch_size=eval_batch_size,
shuffle=False,
num_workers=args.num_workers,
pin_memory=True)
if args.TDE:
causal_embed = np.load(
ckp_path +
'/causal_embed_run_{}_iteration_{}.npy'.format(
iteration_total, iteration))
causal_embed = torch.from_numpy(causal_embed).to(device)
causal_embed = causal_embed / (torch.norm(
causal_embed, 2, 1, keepdim=True))
ori_acc = compute_accuracy_TDE(tg_model,
tg_feature_model,
current_means,
evalloader,
causal_embed,
iteration=iteration -
start_iter)
else:
ori_acc = compute_accuracy(tg_model, tg_feature_model,
current_means, evalloader)
##############################################################
# Calculate validation error of model on the cumul of classes:
map_Y_valid_cumul = np.array(
[order_list.index(i) for i in Y_valid_cumul])
print('Computing cumulative accuracy...')
# evalset.test_data = X_valid_cumul.astype('uint8')
# evalset.test_labels = map_Y_valid_cumul
current_eval_set = merge_images_labels(X_valid_cumul,