def test_gen_model_input(self):
# normal X
result = data_setup.gen_model_input(
data_setup.gen_dataset(self.x, "target", np.mean)["na_filled_dataset"],
n_timesteps=2,
)
self.assertTrue(
(
result[0] == np.array([[[1, 4], [2, 5]], [[2, 5], [3, 6]]], np.float64)
).all()
)
# missing ys
result = data_setup.gen_model_input(
data_setup.gen_dataset(self.missing_y, "target", np.nanmedian)[
"na_filled_dataset"
],
n_timesteps=2,
)
self.assertTrue(
(
result[0]
== np.array(
[
[[1, 4], [2, 7.5]],
[[3, 6], [4, 7]],
[[4, 7], [5, 8]],
[[6, 9], [7, 10]],
],
np.float64,
)
# result[1] == np.array([8, 10, 11, 13], np.float64)
).all()
)
def test_gen_ragged_X(self):
ragged_dataset = data_setup.gen_dataset(self.long_x, "target")[
"for_ragged_dataset"
]
X = data_setup.gen_model_input(ragged_dataset, n_timesteps=2)
result = data_setup.gen_ragged_X(
X[0],
[1, 2],
0,
ragged_dataset,
"target",
fill_ragged_edges=np.nanmedian,
backup_fill_method=np.nanmedian,
)
self.assertTrue(
(
result
== np.array(
[
[[1.0, 7.5], [4.0, 7.5]],
[[2.0, 7.5], [4.0, 7.5]],
[[3.0, 7.5], [4.0, 7.5]],
[[4.0, 7.5], [4.0, 7.5]],
[[5.0, 7.5], [4.0, 7.5]],
[[6.0, 7.5], [4.0, 7.5]],
]
)
).all()
)
def test_instantiate_model(self):
model_input = data_setup.gen_model_input(
data_setup.gen_dataset(self.x, "target")["na_filled_dataset"],
n_timesteps=2)[0] # first tuple of the function, X
result = modelling.instantiate_model(model_input, n_timesteps=2)
self.assertEqual(result["mv_lstm"].n_layers, 2)
self.assertEqual(str(type(result["criterion"])),
"<class 'torch.nn.modules.loss.L1Loss'>")
self.assertEqual(str(type(result["optimizer"])),
"<class 'torch.optim.adam.Adam'>")
def test_predict(self):
model_input = data_setup.gen_model_input(data_setup.gen_dataset(
self.x, "target")["na_filled_dataset"],
n_timesteps=2)
result = modelling.instantiate_model(model_input[0], n_timesteps=2)
model = result["mv_lstm"]
crit = result["criterion"]
opt = result["optimizer"]
model_result = modelling.train_model(model_input[0],
model_input[1],
model,
crit,
opt,
quiet=True)
model = model_result["mv_lstm"]
preds = modelling.predict(model_input[0], model)
self.assertEqual(len(preds), 2)
def test_different_optimizer(self):
model_input = data_setup.gen_model_input(
data_setup.gen_dataset(self.x, "target")["na_filled_dataset"],
n_timesteps=2)[0] # first tuple of the function, X
result = modelling.instantiate_model(model_input,
n_timesteps=2,
optimizer=torch.optim.SGD,
optimizer_parameters={
"lr": 1e-3,
"weight_decay": 0.96
})
self.assertEqual(result["mv_lstm"].n_layers, 2)
self.assertEqual(str(type(result["criterion"])),
"<class 'torch.nn.modules.loss.L1Loss'>")
self.assertEqual(str(type(result["optimizer"])),
"<class 'torch.optim.sgd.SGD'>")
self.assertEqual(
str((result["optimizer"])).strip('\n'),
"SGD (\nParameter Group 0\n dampening: 0\n lr: 0.001\n momentum: 0\n nesterov: False\n weight_decay: 0.96\n)"
.strip('\n'))