def _run_returnn_standalone_net_dict(self):
print(">>> Constructing RETURNN model, load TF checkpoint, run...")
with tf.compat.v1.Session() as session:
from returnn.config import Config
from returnn.tf.network import TFNetwork
config = Config({
"extern_data": {
"data": self._returnn_in_data_dict
},
"debug_print_layer_output_template": True,
})
network = TFNetwork(config=config, name="root")
network.construct_from_dict(self._returnn_net_dict)
network.load_params_from_file(
filename=self._tf_checkpoint_save_path, session=session)
x = network.extern_data.get_default_input_data()
y = network.get_default_output_layer().output
feed_dict = self._make_tf_feed_dict(x)
y_, y_size = session.run((y.placeholder, y.size_placeholder),
feed_dict=feed_dict)
assert isinstance(y_, numpy.ndarray)
print("Output shape:", y_.shape)
numpy.testing.assert_allclose(self._out_returnn_np, y_)
print(">>>> Looks good!")
print()
def _run_returnn_standalone_python(self):
print(
">>> Constructing RETURNN model via Python code, load TF checkpoint, run..."
)
with tf.compat.v1.Session() as session:
with Naming.make_instance(
) as naming: # we expect this to work with the default settings
model_func = self._model_func
# Wrap the model_func in a module.
# We assume this would be flattened away in the namespace.
# All named modules should thus have the same names.
class DummyModule(torch_returnn.nn.Module):
def get_returnn_name(self) -> str:
return "" # also avoid that this name becomes a prefix anywhere
def forward(self, *inputs):
return model_func(wrapped_import_torch_returnn,
*inputs)
dummy_mod = DummyModule()
net_dict = dummy_mod.as_returnn_net_dict(
self._returnn_in_data_dict)
from returnn.config import Config
from returnn.tf.network import TFNetwork
config = Config({
"extern_data": {
"data": self._returnn_in_data_dict
},
"debug_print_layer_output_template": True,
})
network = TFNetwork(config=config, name="root")
network.construct_from_dict(net_dict)
network.load_params_from_file(
filename=self._tf_checkpoint_save_path, session=session)
x = network.extern_data.get_default_input_data()
y = network.get_default_output_layer().output
feed_dict = self._make_tf_feed_dict(x)
y_, y_size = session.run((y.placeholder, y.size_placeholder),
feed_dict=feed_dict)
assert isinstance(y_, numpy.ndarray)
print("Output shape:", y_.shape)
numpy.testing.assert_allclose(self._out_returnn_np, y_)
print(">>>> Looks good!")
print()