def dummy_loss_model(VAR, sea):
AE_path = temp_dir+'DAE_{}_{}.hdf'.format(VAR, sea) # model checkpoint
AE = keras.models.load_model(AE_path)
W = AE.get_weights()
DAE = mu.DAE(AE_N, AE_input_size, AE_latent_size)
DAE.set_weights(W)
# Encoder layer selection
encoder = DAE.layers[1]
# freeze enoder layers
encoder.trainable = False
for layer in encoder.layers:
layer.trainable = False
f_sproj = [encoder.layers[i].output for i in layer_id]
# Loss models
loss_models = []
for single_proj in f_sproj:
loss_models.append(keras.models.Model(encoder.inputs, single_proj))
return loss_models
def loss_model(VAR, sea, layer_id):
model_path = temp_dir + 'DAE_{}_{}_self.hdf'.format(
VAR, sea) # model checkpoint
AE = keras.models.load_model(model_path)
W = AE.get_weights()
N = [48, 96, 192, 384]
input_size = (None, None, 1)
# DAE
DAE = mu.DAE(N, input_size)
DAE.set_weights(W)
# freeze
DAE.trainable = False
for layer in DAE.layers:
layer.trainable = False
f_sproj = [DAE.layers[i].output for i in layer_id]
loss_models = []
for single_proj in f_sproj:
loss_models.append(keras.models.Model(DAE.inputs, single_proj))
return loss_models
l = [1e-4, 1e-5] # learning rate
epochs = 200 # sapce for early stopping
# DAE
N = [48, 96, 192, 384]
input_size = (None, None, 1)
# training file location
file_path = BATCH_dir
trainfiles = glob(file_path+'{}_BATCH_*_TORI*_{}*.npy'.format(VAR, sea))
validfiles = glob(file_path+'{}_BATCH_*_VORI*_{}*.npy'.format(VAR, sea))
#
model_path = temp_dir+'DAE_{}_{}_elev.hdf'.format(VAR, sea)
train_path = temp_dir+'DAE_{}_{}_elev.npy'.format(VAR, sea)
DAE = mu.DAE(N, input_size)
# optimizer & callback & compile
opt_ae = keras.optimizers.Adam(lr=l[0])
callbacks = [keras.callbacks.EarlyStopping(monitor='val_loss', min_delta=0.00001, patience=2, verbose=True),
keras.callbacks.ModelCheckpoint(filepath=model_path, verbose=True, monitor='val_loss', save_best_only=True)]
DAE.compile(loss=keras.losses.mean_absolute_error, optimizer=opt_ae, metrics=[keras.losses.mean_absolute_error])
# Data generator
gen_train = tu.grid_grid_gen(trainfiles, labels, input_flag, output_flag)
gen_valid = tu.grid_grid_gen(validfiles, labels, input_flag, output_flag)
# train
temp_hist = DAE.fit_generator(generator=gen_train, validation_data=gen_valid, callbacks=callbacks,
initial_epoch=0, epochs=epochs, verbose=1, shuffle=True, max_queue_size=8, workers=8)