In my architecture, i have to reuse a module with the same weights as its original ones. Does the following method is the right way to realize my thought?
Class Encoder(nn.module):
def init(self):
super(Encoder, self).init()
。。。
def forward(self, x1, x2, emb, input_sel):
if input_sel == 0 :
x = x1
else:
x = x2
x = 。。。。
。。。
Class Top(nn.module):
def init(self):
super(Top, self).init()
self.encoder = Encoder()
self.decoder = Decoder()
。。。
def forward(input, emb1,emb2):
encoder_output = self.encoder(input,input,emb1,0)
decoder_output = self.decoder(encoder_output)
code_output = self.encoder(decoder_output,decoder_output,emb1,1)
return encoder_output, decoder_output, code_output
Can the two encoders share the same weight in above method??