def __init__(self, params: Dict[str, Any]):
"""Initialise the layer."""
super().__init__()
self._params = params
self._hidden_dim = params["hidden_dim"]
self._num_layers = params["num_layers"]
self._dense_every_num_layers = params["dense_every_num_layers"]
self._residual_every_num_layers = params["residual_every_num_layers"]
self._use_inter_layer_layernorm = params["use_inter_layer_layernorm"]
self._initial_node_representation_activation_fn = get_activation_function(
params["initial_node_representation_activation"]
)
self._dense_intermediate_layer_activation_fn = get_activation_function(
params["dense_intermediate_layer_activation"]
)
self._message_passing_class = get_message_passing_class(
params["message_calculation_class"]
)
if not params["global_exchange_mode"].lower() in {"mean", "mlp", "gru"}:
raise ValueError(
f"Unknown global_exchange_mode mode {params['global_exchange_mode']} - has to be one of 'mean', 'mlp', 'gru'!"
)
self._global_exchange_mode = params["global_exchange_mode"]
self._global_exchange_every_num_layers = params["global_exchange_every_num_layers"]
self._global_exchange_weighting_fun = params["global_exchange_weighting_fun"]
self._global_exchange_num_heads = params["global_exchange_num_heads"]
self._global_exchange_dropout_rate = params["global_exchange_dropout_rate"]
# Layer member variables. To be filled in in the `build` method.
self._initial_projection_layer: tf.keras.layers.Layer = None
self._mp_layers: List[MessagePassing] = []
self._inter_layer_layernorms: List[tf.keras.layers.Layer] = []
self._dense_layers: Dict[str, tf.keras.layers.Layer] = {}
self._global_exchange_layers: Dict[str, GraphGlobalExchange] = {}
def __init__(self, params: Dict[str, Any], **kwargs):
super().__init__(**kwargs)
self._hidden_dim = int(params["hidden_dim"])
aggregation_fn_name = params["aggregation_function"]
self._aggregation_fn = get_aggregation_function(aggregation_fn_name)
activation_fn_name = params["message_activation_function"]
self._activation_fn = get_activation_function(activation_fn_name)
def __init__(self, params: Dict[str, Any], **kwargs):
super().__init__(**kwargs)
self._hidden_dim = int(params["hidden_dim"])
aggregation_fn_name = params["aggregation_function"]
self._aggregation_fn = get_aggregation_function(aggregation_fn_name)
activation_fn_name = params["message_activation_function"]
self._activation_fn = get_activation_function(activation_fn_name)
self._hyperedge_type_mlps: List[tf.keras.layers.Layer, ...] = []
self._num_edge_MLP_hidden_layers = params["num_edge_MLP_hidden_layers"]