def test_train_05(self):
nll = NonlinearLevelSet(n_layers=2, active_dim=1, lr=0.02, epochs=1)
with assert_plot_figures_added():
nll.train(inputs=inputs_torch,
gradients=grad_torch,
outputs=lift,
interactive=True)
def test_save_backward(self):
nll = NonlinearLevelSet(n_layers=2, active_dim=1, lr=0.02, epochs=1)
nll.train(inputs=inputs_torch, gradients=grad_torch, interactive=False)
outfilename = 'tests/data/saved_backward.pth'
nll.save_backward(outfilename)
nll.save_backward('tests/data/backward_test.pth')
self.assertTrue(os.path.exists(outfilename))
self.addCleanup(os.remove, outfilename)
def test_init_loss_vec(self):
nll = NonlinearLevelSet(n_layers=2,
active_dim=1,
lr=0.1,
epochs=100,
dh=0.25)
self.assertEqual(nll.loss_vec, [])
def test_init_backward(self):
nll = NonlinearLevelSet(n_layers=2,
active_dim=1,
lr=0.1,
epochs=100,
dh=0.25)
self.assertIsNone(nll.backward)
def test_load_backward(self):
nll = NonlinearLevelSet(n_layers=2, active_dim=1, lr=0.02, epochs=1)
nll.load_backward(infile='tests/data/backward_test.pth', n_params=18)
self.assertIsInstance(nll.backward, BackwardNet)
def test_plot_loss(self):
nll = NonlinearLevelSet(n_layers=2, active_dim=1, lr=0.02, epochs=2)
nll.train(inputs=inputs_torch, gradients=grad_torch, interactive=False)
with assert_plot_figures_added():
nll.plot_loss()
def test_plot_sufficient_summary_02(self):
nll = NonlinearLevelSet(n_layers=2, active_dim=2, lr=0.02, epochs=1)
nll.train(inputs=inputs_torch, gradients=grad_torch, interactive=False)
with self.assertRaises(ValueError):
nll.plot_sufficient_summary(inputs=inputs_torch, outputs=lift)
def test_plot_sufficient_summary_01(self):
nll = NonlinearLevelSet(n_layers=2, active_dim=1, lr=0.02, epochs=1)
nll.train(inputs=inputs_torch, gradients=grad_torch, interactive=False)
with assert_plot_figures_added():
nll.plot_sufficient_summary(inputs=inputs_torch, outputs=lift)
def test_backward_n_params(self):
nll = NonlinearLevelSet(n_layers=2, active_dim=1, lr=0.02, epochs=1)
nll.train(inputs=inputs_torch, gradients=grad_torch, interactive=False)
self.assertEqual(nll.backward.n_params, 9)
def test_train_04(self):
nll = NonlinearLevelSet(n_layers=2, active_dim=1, lr=0.02, epochs=1)
with self.assertRaises(ValueError):
nll.train(inputs=inputs_torch,
gradients=grad_torch,
interactive=True)
def test_train_03(self):
nll = NonlinearLevelSet(n_layers=2, active_dim=1, lr=0.02, epochs=1)
nll.train(inputs=inputs_torch, gradients=grad_torch, interactive=False)
self.assertIs(len(nll.loss_vec), 1)
def test_train_02(self):
nll = NonlinearLevelSet(n_layers=2, active_dim=1, lr=0.02, epochs=1)
nll.train(inputs=inputs_torch, gradients=grad_torch, interactive=False)
self.assertIsInstance(nll.backward, BackwardNet)
