" | CURRICULUM LEARNING: " + str(args.curriculum_learning) +
" | MODEL: " + str(args.type_model))
print()
nn_model_ref = NN_Model_Ref(args, writer, device, rng, path_save_model,
cipher, creator_data_binary,
path_save_model_train)
if args.retain_model_gohr_ref:
nn_model_ref.train_general(name_input)
else:
#nn_model_ref.load_nn()
try:
if args.finetunning:
nn_model_ref.load_nn()
nn_model_ref.train_from_scractch(name_input + "fine-tune")
#nn_model_ref.eval(["val"])
else:
nn_model_ref.load_nn()
except:
print("ERROR")
print("NO MODEL AVALAIBLE FOR THIS CONFIG")
print("CHANGE ARGUMENT retain_model_gohr_ref")
print()
sys.exit(1)
if args.create_new_data_for_ToT and args.create_new_data_for_classifier:
del nn_model_ref.X_train_nn_binaire, nn_model_ref.X_val_nn_binaire, nn_model_ref.Y_train_nn_binaire, nn_model_ref.Y_val_nn_binaire
del nn_model_ref.c0l_train_nn, nn_model_ref.c0l_val_nn, nn_model_ref.c0r_train_nn, nn_model_ref.c0r_val_nn
del nn_model_ref.c1l_train_nn, nn_model_ref.c1l_val_nn, nn_model_ref.c1r_train_nn, nn_model_ref.c1r_val_nn
net = NN_linear(args, X_train_f.shape[1]).to(device)
nn_model_ref.net = net
nn_model_ref.X_train_nn_binaire = X_train_f
nn_model_ref.X_val_nn_binaire = X_val_f
#nn_model_ref.Y_train_nn_binaire = X_train_f
#nn_model_ref.Y_val_nn_binaire = X_val_f
"""
args.load_nn_path = "./results/create_synth_masks_v2/speck/5/ctdata0l^ctdata1l_ctdata0r^ctdata1r^ctdata0l^ctdata1l_ctdata0l^ctdata0r_ctdata1l^ctdata1r/2020_07_21_17_26_59_603174/0.9966710913033485_bestacc.pth"
nn_model_ref.net.load_state_dict(torch.load(args.load_nn_path,
map_location=device)['state_dict'], strict=False)
nn_model_ref.net.to(device)
nn_model_ref.net.eval()
"""
nn_model_ref.train_from_scractch("AE")
for global_sparsity in [0, 0.2, 0.4]:
print(global_sparsity)
flag2 = True
acc_retain = []
parameters_to_prune = []
for name, module in nn_model_ref.net.named_modules():
if len(name):
if name not in ["layers_batch", "layers_conv"]:
flag = True
for layer_forbidden in args.layers_NOT_to_prune:
if layer_forbidden in name:
flag = False
if flag:
parameters_to_prune.append((module, 'weight'))