def test_model_q(model_name='q', region="50S69W"):
"""
Test the humidity model
"""
global model
# Load testing data
train_data_in, train_data_out, test_data_in, test_data_out = data_io.scm_model_data(region)
q_norm_test = test_data_in["qtot_test"]
qadv_norm_test = test_data_in["qadv_test"]
qadv_dot_norm_test = test_data_in["qadv_dot_test"]
qphys_norm_test = test_data_out["qphys_test"]
qadd_test = test_data_in["qadd_test"]
qadd_dot_test = test_data_in["qadd_dot_test"]
qcomb_test = np.concatenate((qadv_norm_test,q_norm_test),axis=1)
qcomb_dot_test = np.concatenate((qadd_dot_test,q_norm_test),axis=1)
# qcomb_dot_test = np.concatenate((qadv_dot_norm_test,q_norm_test),axis=1)
checkpoint_dir = '{0}/{1}'.format(locations['chkpnt_loc'], model_name)
latest = tf.train.latest_checkpoint(checkpoint_dir)
# After compiling model, run through some data for initialising variables then load weights
model.compile(optimizer=optimizer,loss=loss_object,metrics=[test_accuracy])
# model(qadv_test[:1])
# model(q_test[:1])
# model(qcomb_test[:1])
model(qcomb_dot_test[:1])
model.load_weights(latest)
print(model_name)
print(model.summary())
qphys_normaliser = joblib.load('{0}/minmax_qphystot.joblib'.format(locations['normaliser_loc']))
q_normaliser = joblib.load('{0}/minmax_qtot.joblib'.format(locations['normaliser_loc']))
qadv_normaliser = joblib.load('{0}/minmax_qadvtot.joblib'.format(locations['normaliser_loc']))
qadv_dot_normaliser = joblib.load('{0}/minmax_qadv_dot.joblib'.format(locations['normaliser_loc']))
# qphys_predict = model.predict(qadv_test)
# qphys_predict = model.predict(q_test)
# qphys_predict = model.predict(qcomb_test)
qadv_predict = model.predict(qcomb_dot_test)
# qphys_predict_denorm = qphys_normaliser.inverse_transform(qphys_predict)
# qphys_test_denorm = qphys_normaliser.inverse_transform(qphys_test)
qadv_predict_denorm = qadv_normaliser.inverse_transform(qadv_predict)
qadv_test_denorm = qadv_normaliser.inverse_transform(qadv_norm_test)
# np.savez('qtot_predict',qphys_predict=qphys_predict_denorm,qphys_test=qphys_test_denorm)
np.savez('qcomb_dot_predict_qadv',qadv_predict=qadv_predict_denorm,qadv_test=qadv_test_denorm)
locations = {
"train_test_datadir": "/project/spice/radiation/ML/CRM/data/models/datain",
"chkpnt_loc": "/project/spice/radiation/ML/CRM/data/models/chkpts",
"pkl_loc": "/project/spice/radiation/ML/CRM/data/models/history",
"normaliser_loc": "/project/spice/radiation/ML/CRM/data/models/normaliser"
}
if __name__ == "__main__":
BATCH = 100
EPOCHS = 50
# model_name = "qcomb_add_dot_qphys_deep"
model_name = "tcomb_add_dot_qphys_deep"
region = "50S69W"
train_data_in, train_data_out, test_data_in, test_data_out = data_io.scm_model_data(
region)
t_norm_train = train_data_in["T"]
tadv_norm_train = train_data_in["tadv"]
tadv_dot_norm_train = train_data_in["tadv_dot"]
tphys_norm_train = train_data_out["tphys_tot"]
t_norm_test = test_data_in["T_test"]
tadv_norm_test = test_data_in["tadv_test"]
tadv_dot_norm_test = test_data_in["tadv_dot_test"]
tphys_norm_test = test_data_out["tphys_test"]
tadd_train = train_data_in["tadd"]
tadd_dot_train = train_data_in["tadd_dot"]
tadd_test = test_data_in["tadd_test"]
tadd_dot_test = test_data_in["tadd_dot_test"]
tcomb_train = np.concatenate((tadv_norm_train, t_norm_train), axis=1)
tcomb_test = np.concatenate((tadv_norm_test, t_norm_test), axis=1)
loss_object = tf.keras.losses.MeanSquaredError()
optimizer = tf.keras.optimizers.Adam()
#optimizer = tf.keras.optimizers.SGD(lr=0.1)
# tf.print(optimizer.learning_rate, output_stream=sys.stdout)
train_loss = tf.keras.metrics.MeanSquaredError(name='train_loss')
train_accuracy = tf.keras.metrics.CategoricalAccuracy(name='train_accuracy')
test_loss = tf.keras.metrics.MeanSquaredError(name='test_loss')
test_accuracy = tf.keras.metrics.CategoricalAccuracy(name='test_accuracy')
locations = {
"train_test_datadir": "/project/spice/radiation/ML/CRM/data/models/datain",
"chkpnt_loc": "/project/spice/radiation/ML/CRM/data/models/chkpts",
"pkl_loc": "/project/spice/radiation/ML/CRM/data/models/history"
}
if __name__ == "__main__":
BATCH = 100
EPOCHS = 50
model_name = "scm"
region = "50S69W"
model_data_in_train, model_data_out_train, model_data_in_test, model_data_out_test = data_io.scm_model_data(
region)
train_data = []
test_data = []
create_train_scm_model(BATCH,
EPOCHS,
train_data,
test_data,
model_name=model_name,
region=region)