5ad1bd651f39457efcebe6be251204a4cc8f07e8,theanolm/network.py,Architecture,from_state,#Any#Any#,47

Before Change

            classname._check_parameter_in_state("arch.hidden_layer_size", state)
            classname._check_parameter_in_state("arch.skip_layer_size", state)
            return classname(
                state["arch.word_projection_dim"].item(),
                state["arch.hidden_layer_type"].item(),
                state["arch.hidden_layer_size"].item(),
                state["arch.skip_layer_size"].item())

After Change

            :param state: a dictionary of the architecture parameters
            

            if not "arch.layers" in state:
                raise IncompatibleStateError(
                    "Parameter "arch.layers" is missing from neural network state.")
            // A workaround to be able to save arbitrary data in a .npz file.
            dict_ndarray = state["arch.layers"][()]
            layer_descriptions = dict_ndarray["data"]
            return classname(layer_descriptions)

Instances