def solver(generators):
return GenericSolver(
diff_eqs=DIFF_EQS,
conditions=CONDITIONS,
train_generator=generators['train'],
valid_generator=generators['valid'],
n_input_units=1,
n_output_units=1,
)
def test_legacies(solver, generators):
solver.fit(1)
assert solver.batch == solver._batch
with pytest.warns(FutureWarning):
assert solver._batch_examples == solver._batch
with pytest.raises(TypeError), pytest.warns(FutureWarning):
GenericSolver(
diff_eqs=DIFF_EQS,
conditions=CONDITIONS,
train_generator=generators['train'],
valid_generator=generators['valid'],
criterion=lambda residuals, zeros: (residuals ** 2).mean(),
n_input_units=1,
n_output_units=1,
).fit(1)
class SolverWithLegacyAdditionalLoss(BaseSolver):
def additional_loss(self, funcs, key):
return 0
with pytest.raises(TypeError), pytest.warns(FutureWarning):
SolverWithLegacyAdditionalLoss(
diff_eqs=DIFF_EQS,
conditions=CONDITIONS,
train_generator=generators['train'],
valid_generator=generators['valid'],
n_input_units=1,
n_output_units=1,
).fit(1)
with pytest.warns(FutureWarning):
GenericSolver(
diff_eqs=DIFF_EQS,
conditions=CONDITIONS,
train_generator=generators['train'],
valid_generator=generators['valid'],
n_input_units=1,
n_output_units=1,
shuffle=True,
)
def test_lbfgs(generators):
nets = [FCNN()]
lbfgs = torch.optim.LBFGS(params=nets[0].parameters(), lr=1e-3)
GenericSolver(
diff_eqs=DIFF_EQS,
conditions=CONDITIONS,
train_generator=generators['train'],
valid_generator=generators['valid'],
nets=nets,
optimizer=lbfgs,
n_input_units=1,
n_output_units=1,
).fit(1)
def test_missing_generator(generators):
with pytest.raises(ValueError):
GenericSolver(
diff_eqs=DIFF_EQS,
conditions=CONDITIONS,
train_generator=generators['train'],
n_input_units=1,
n_output_units=1,
)
with pytest.raises(ValueError):
GenericSolver(
diff_eqs=DIFF_EQS,
conditions=CONDITIONS,
valid_generator=generators['valid'],
n_input_units=1,
n_output_units=1,
)
with pytest.raises(ValueError):
GenericSolver(
diff_eqs=DIFF_EQS,
conditions=CONDITIONS,
n_input_units=1,
n_output_units=1,
)