def test_save_tf(self):
path = os.path.join(self.get_temp_dir(), 'model')
with self.assertRaisesRegexp(
NotImplementedError,
'Saving the model as SavedModel is still in experimental stages.'
):
save.save_model(self.model, path, save_format='tf')
def test_save_format_defaults_pathlib(self):
if sys.version_info < (3, 6):
self.skipTest(
'pathlib is only available for python version >= 3.6')
path = pathlib.Path(self.get_temp_dir()) / 'model_path'
save.save_model(self.model, path)
self.assert_saved_model(path)
def test_save_load_hdf5_pathlib(self):
if sys.version_info < (3, 6):
self.skipTest(
'pathlib is only available for python version >= 3.6')
path = pathlib.Path(self.get_temp_dir()) / 'model'
save.save_model(self.model, path, save_format='h5')
save.load_model(path)
def save_model(model, path, optimizer=True):
if isinstance(path, str):
if path.endswith(".h5") or path.endswith(".hdf5"):
filepath = path
else:
raise ValueError("file must be .h5 file")
else:
raise TypeError("file must be type `str`")
save.save_model(model.train_model, path, include_optimizer=optimizer)
with h5py.File(filepath, "r+") as h5file:
train_generator = model.train_generator
h5file.attrs["train_generator_config"] = json.dumps(
{
"class_name": train_generator.__class__.__name__,
"config": train_generator.get_config(),
},
default=get_json_type,
).encode("utf8")
h5file.attrs["pose_model_config"] = json.dumps(
{
"class_name": model.__class__.__name__,
"config": model.get_config()
},
default=get_json_type,
).encode("utf8")
def test_save_hdf5(self):
path = os.path.join(self.get_temp_dir(), 'model')
save.save_model(self.model, path, save_format='h5')
self.assert_h5_format(path)
with self.assertRaisesRegex(
NotImplementedError,
'requires the model to be a Functional model or a Sequential model.'):
save.save_model(self.subclassed_model, path, save_format='h5')
def test_layer_saving_with_h5(self):
vocab_data = ["earth", "wind", "and", "fire"]
input_data = keras.Input(shape=(None, ), dtype=dtypes.string)
layer = get_layer_class()(max_tokens=10)
layer.set_vocabulary(vocab_data)
int_data = layer(input_data)
model = keras.Model(inputs=input_data, outputs=int_data)
path = os.path.join(self.get_temp_dir(), "model")
with self.assertRaisesRegex(NotImplementedError,
"Save or restore weights that is not.*"):
save.save_model(model, path, save_format="h5")
def test_load_weights(self):
"""
test loading model weights
"""
model = testing_utils.get_small_functional_mlp(
3, 2, 3, ["fc", "output", "input"])
model.compile(loss=keras.losses.MSE,
optimizer=keras.optimizers.RMSprop(lr=0.0001),
metrics=[keras.metrics.categorical_accuracy])
path = os.path.join(self.get_temp_dir(), 'model')
save.save_model(model, path, save_format='h5')
del model
model = testing_utils.get_small_functional_mlp(
2, 2, 3, ["fc", "output", "input"])
model.load_weights(path, by_name=True, skip_mismatch=True)
with self.assertRaises(ValueError):
model.load_weights(path, by_name=True, skip_mismatch=False)
del model
def test_save_tf(self):
path = os.path.join(self.get_temp_dir(), 'model')
save.save_model(self.model, path, save_format='tf')
self.assert_saved_model(path)
with self.assertRaisesRegex(ValueError, 'input shapes have not been set'):
save.save_model(self.subclassed_model, path, save_format='tf')
self.subclassed_model.predict(np.random.random((3, 5)))
save.save_model(self.subclassed_model, path, save_format='tf')
self.assert_saved_model(path)
def test_save_format_defaults(self): path = os.path.join(self.get_temp_dir(), 'model_path') save.save_model(self.model, path) self.assert_saved_model(path)
def test_save_load_tf_string(self): path = os.path.join(self.get_temp_dir(), 'model') save.save_model(self.model, path, save_format='tf') save.load_model(path)
def test_save_load_tf_pathlib(self):
if sys.version_info >= (3, 4):
path = pathlib.Path(self.get_temp_dir()) / 'model'
save.save_model(self.model, path, save_format='tf')
save.load_model(path)
def test_save_format_defaults(self):
path = os.path.join(self.get_temp_dir(), 'model_path')
# The default is currently HDF5 no matter what the filepath is.
save.save_model(self.model, path)
self.assert_h5_format(path)
def test_save_hdf5(self):
path = os.path.join(self.get_temp_dir(), 'model')
save.save_model(self.model, path, save_format='h5')
self.assert_h5_format(path)
def test_save_format_defaults_pathlib(self): path = pathlib.Path(self.get_temp_dir()) / 'model_path' save.save_model(self.model, path) self.assert_saved_model(path)
def test_save_format_defaults(self):
path = os.path.join(self.get_temp_dir(), 'model_path')
# The default is currently HDF5 no matter what the filepath is.
save.save_model(self.model, path)
self.assert_h5_format(path)
def test_save_tf(self):
path = os.path.join(self.get_temp_dir(), 'model')
with self.assertRaisesRegexp(
NotImplementedError,
'Saving the model as SavedModel is still in experimental stages.'):
save.save_model(self.model, path, save_format='tf')
def test_save_hdf5(self):
path = os.path.join(self.get_temp_dir(), 'model')
save.save_model(self.model, path, save_format='h5')
self.assert_h5_format(path)
def _save_model(self, model, saved_dir):
save.save_model(model, saved_dir, save_format='tf')
def test_save_load_hdf5_pathlib(self): path = pathlib.Path(self.get_temp_dir()) / 'model' save.save_model(self.model, path, save_format='h5') save.load_model(path)