def _check_model_compatibility(self, y: np.ndarray) -> None:
"""Checks that the model output number and y shape match.
This is in place to avoid cryptic TF errors.
"""
# check if this is a multi-output model
if hasattr(self, "n_outputs_expected_"):
# n_outputs_expected_ is generated by data transformers
# we recognize the attribute but do not force it to be
# generated
if self.n_outputs_expected_ != len(self.model_.outputs):
raise ValueError("Detected a Keras model input of size"
f" {y[0].shape[0]}, but {self.model_} has"
f" {self.model_.outputs} outputs")
# check that if the user gave us a loss function it ended up in
# the actual model
init_params = inspect.signature(self.__init__).parameters
if "loss" in init_params:
default_val = init_params["loss"].default
if all(
isinstance(x, (str, losses_module.Loss, type))
for x in [self.loss, self.model_.loss
]): # filter out loss list/dicts/etc.
if default_val is not None:
default_val = loss_name(default_val)
given = loss_name(self.loss)
got = loss_name(self.model_.loss)
if given != default_val and got != given:
raise ValueError(
f"loss={self.loss} but model compiled with {self.model_.loss}."
" Data may not match loss function!")
def test_unknown_loss_raises():
with pytest.raises(ValueError, match="Unknown loss function"):
loss_name("unknown_loss")
def test_loss_types(loss):
with pytest.raises(TypeError, match="`loss` must be a"):
loss_name(loss)
def test_custom_loss(obj):
assert loss_name(obj) == "custom_loss"
def test_loss_invariance(obj):
"""Test to make sure loss_name returns same string no matter which object
is passed (str, function, class, type)"""
assert loss_name(obj) == "categorical_crossentropy"