def test_save_load():
"""Test the save/load operations of TfModel"""
hidden_layers = [{
'units': 64,
'activation': 'relu',
'name': 'relu1'
}, {
'units': 64,
'activation': 'relu',
'name': 'relu2'
}]
model = TfModel.build_trained(FEATURES,
LABELS,
hidden_layers=hidden_layers,
epochs=10,
fit_kwargs={"batch_size": 16},
early_stop=False,
save_path=FPATH)
y_pred = model[X]
loaded = TfModel.load(FPATH)
y_pred_loaded = loaded[X]
np.allclose(y_pred.values, y_pred_loaded.values)
assert loaded.feature_names == ['a', 'b']
assert loaded.label_names == ['c']
shutil.rmtree(FPATH)
os.remove(FPATH_JSON)
def test_bad_categories():
"""
Test OHE checks
"""
hidden_layers = [{'units': 64, 'activation': 'relu', 'name': 'relu1'},
{'units': 64, 'activation': 'relu', 'name': 'relu2'}]
one_hot_categories = {'categorical': list('abc')}
feature_names = FEATURES.columns.tolist() + ['categorical']
label_names = 'c'
with pytest.raises(RuntimeError):
TfModel.build(feature_names, label_names,
one_hot_categories=one_hot_categories,
hidden_layers=hidden_layers)
one_hot_categories = {'categorical': list('cdf')}
feature_names = FEATURES.columns.tolist() + ['categorical']
label_names = 'c'
with pytest.raises(RuntimeError):
TfModel.build(feature_names, label_names,
one_hot_categories=one_hot_categories,
hidden_layers=hidden_layers)
one_hot_categories = {'categorical': list('def')}
feature_names = FEATURES.columns.tolist() + ['categories']
label_names = 'c'
with pytest.raises(RuntimeError):
TfModel.build(feature_names, label_names,
one_hot_categories=one_hot_categories,
hidden_layers=hidden_layers)
one_hot_categories = {'cat1': list('def'), 'cat2': list('fgh')}
feature_names = FEATURES.columns.tolist() + ['cat1', 'cat2']
label_names = 'c'
with pytest.raises(RuntimeError):
TfModel.build(feature_names, label_names,
one_hot_categories=one_hot_categories,
hidden_layers=hidden_layers)
ohe_features = FEATURES.copy()
categories = list('def')
ohe_features['categorical'] = np.random.choice(categories, len(FEATURES))
one_hot_categories = {'categorical': categories}
model = TfModel.build_trained(ohe_features, LABELS,
one_hot_categories=one_hot_categories,
hidden_layers=hidden_layers,
epochs=10, fit_kwargs={"batch_size": 16},
early_stop=False)
with pytest.raises(RuntimeError):
x = ohe_features.values[:, 1:]
model.predict(x)
def test_complex_nn():
"""Test complex TfModel """
hidden_layers = [{
'units': 64,
'activation': 'relu',
'dropout': 0.01
}, {
'units': 64
}, {
'batch_normalization': {
'axis': -1
}
}, {
'activation': 'relu'
}, {
'dropout': 0.01
}]
model = TfModel.build_trained(FEATURES,
LABELS,
hidden_layers=hidden_layers,
epochs=10,
fit_kwargs={"batch_size": 16},
early_stop=False)
assert len(model.layers) == 8
assert len(model.weights) == 10
test_mae = np.mean(np.abs(model[X].values - Y))
loss = 0.15
assert model.history['val_mae'].values[-1] < loss
assert test_mae < loss
def test_OHE():
"""
Test one-hot encoding
"""
ohe_features = FEATURES.copy()
categories = list('def')
ohe_features['categorical'] = np.random.choice(categories, len(FEATURES))
one_hot_categories = {'categorical': categories}
hidden_layers = [{
'units': 64,
'activation': 'relu',
'name': 'relu1'
}, {
'units': 64,
'activation': 'relu',
'name': 'relu2'
}]
model = TfModel.build_trained(ohe_features,
LABELS,
one_hot_categories=one_hot_categories,
hidden_layers=hidden_layers,
epochs=10,
fit_kwargs={"batch_size": 16},
early_stop=False)
assert all(np.isin(categories, model.feature_names))
assert not any(np.isin(categories, model.input_feature_names))
assert 'categorical' not in model.feature_names
assert 'categorical' in model.input_feature_names
x = ohe_features.values
out = model.predict(x)
assert 'c' in out
def test_save_load():
"""Test the save/load operations of TfModel"""
with tempfile.TemporaryDirectory() as td:
model_fpath = os.path.join(td, 'test_model/')
hidden_layers = [{'units': 64, 'activation': 'relu', 'name': 'relu1'},
{'units': 64, 'activation': 'relu', 'name': 'relu2'}]
model = TfModel.build_trained(FEATURES, LABELS,
hidden_layers=hidden_layers,
epochs=10, fit_kwargs={"batch_size": 16},
early_stop=False,
save_path=model_fpath)
y_pred = model[X]
loaded = TfModel.load(model_fpath)
y_pred_loaded = loaded[X]
np.allclose(y_pred.values, y_pred_loaded.values)
assert loaded.feature_names == ['a', 'b']
assert loaded.label_names == ['c']
def test_normalize(normalize, loss):
"""Test TfModel """
hidden_layers = [{'units': 64, 'activation': 'relu', 'name': 'relu1'},
{'units': 64, 'activation': 'relu', 'name': 'relu2'}]
model = TfModel.build_trained(FEATURES, LABELS,
normalize=normalize,
hidden_layers=hidden_layers,
epochs=10, fit_kwargs={"batch_size": 16},
early_stop=False)
test_mae = np.mean(np.abs(model[X].values - Y))
assert model.history[mae_key].values[-1] < loss
assert test_mae < loss
def test_nn(hidden_layers, loss):
"""Test TfModel """
model = TfModel.build_trained(FEATURES, LABELS,
hidden_layers=hidden_layers,
epochs=10,
fit_kwargs={"batch_size": 16},
early_stop=False)
n_l = len(hidden_layers) * 2 + 1 if hidden_layers is not None else 1
n_w = (len(hidden_layers) + 1) * 2 if hidden_layers is not None else 2
assert len(model.layers) == n_l
assert len(model.weights) == n_w
assert len(model.history) == 10
test_mae = np.mean(np.abs(model[X].values - Y))
assert model.history[mae_key].values[-1] < loss
assert test_mae < loss
def test_dropout():
"""Test a model trained with dropout vs. no dropout and make sure the
predictions are different."""
hidden_layers_1 = [{
'units': 64,
'activation': 'relu'
}, {
'units': 64
}, {
'activation': 'relu'
}]
hidden_layers_2 = [{
'units': 64,
'activation': 'relu',
'dropout': 0.05
}, {
'units': 64
}, {
'activation': 'relu'
}, {
'dropout': 0.05
}]
TfModel.seed()
model_1 = TfModel.build_trained(FEATURES,
LABELS,
hidden_layers=hidden_layers_1,
epochs=10,
fit_kwargs={"batch_size": 16},
early_stop=False)
TfModel.seed()
model_2 = TfModel.build_trained(FEATURES,
LABELS,
hidden_layers=hidden_layers_2,
epochs=10,
fit_kwargs={"batch_size": 16},
early_stop=False)
out1 = model_1.history['val_mae'].values[-5:]
out2 = model_2.history['val_mae'].values[-5:]
assert (out2 > out1).all()
"""
Tests for basic tensorflow model functionality and execution.
"""
# pylint: disable=W0613
import numpy as np
import os
import pandas as pd
import pytest
import tempfile
from phygnn.utilities import TF2
from phygnn.model_interfaces.tf_model import TfModel
TfModel.seed(0)
if TF2:
mae_key = 'val_mae'
else:
mae_key = 'val_mean_absolute_error'
N = 100
A = np.linspace(-1, 1, N)
B = np.linspace(-1, 1, N)
A, B = np.meshgrid(A, B)
A = np.expand_dims(A.flatten(), axis=1)
B = np.expand_dims(B.flatten(), axis=1)
Y = np.sqrt(A ** 2 + B ** 2)
X = np.hstack((A, B))
FEATURES = pd.DataFrame(X, columns=['a', 'b'])