how can I define the 3 #MGCN layers for YAGO as described in table 8 of paper? Bug in MGCN.py?

[JuliaGast]

Dear All,

in table 8 of the paper you describe that for YAGO you set # MGCN layer =3.

As far as I understand, this can be set by the arg core_layer, correct?

However, in the code, when I look at MGCN.py, as far as I understand, the MGCN will not deal with more than two layers (please see lines with comments titled #HERE below:):

		# define MGCN Layer
		self.conv1 = MGCNConvLayer(self.p.initsize, self.p.hidsize, act=self.act, params=self.p)
		self.conv2 = MGCNConvLayer(self.p.hidsize, self.p.embsize, act=self.act, params=self.p) if self.p.core_layer == 2 else None #HERE


		self.register_parameter('bias', Parameter(torch.zeros(num_e)))

	def set_graph(self, edge_index, edge_type):
		self.edge_index = edge_index
		self.edge_type = edge_type

	def set_jumpfunc(self, edge_id_jump, edge_w_jump, jumpfunc, jumpw=None, skip=False, rel_jump=None):
		self.edge_id_jump = edge_id_jump
		self.edge_w_jump = edge_w_jump
		self.jump = jumpfunc
		self.jump_weight = jumpw
		self.skip = skip
		self.rel_jump = rel_jump

	def forward(self, t, emb):
		self.nfe += 1
		jump_emb = emb.clone()
		if self.p.res:
			emb = emb + self.res * self.conv1(emb, self.edge_index, self.edge_type, self.num_e)
			emb = self.drop_l1(emb)
			**emb = (emb + self.res * self.conv2(emb, self.edge_index, self.edge_type, self.num_e)) if self.p.core_layer == 2 else #HERE
 emb**
			**emb = self.drop_l2(emb) if self.p.core_layer == 2 else emb** #HERE

		else:
			emb	= self.conv1(emb, self.edge_index, self.edge_type, self.num_e)
			emb	= self.drop_l1(emb)
			**emb	= self.conv2(emb, self.edge_index, self.edge_type, self.num_e) 	if self.p.core_layer == 2 else emb** #HERE
			**emb	= self.drop_l2(emb) 							if self.p.core_layer == 2 else emb** #HERE

This means, if self.p.core_layer is 3, thus != 2, it will treat this as one layer. Is my understanding correct or am I mistaken?

If I'm correct: Does this mean, that the hyperparameter for YAGO should actually be core_layer=1?

Best and Thanks