Update AlphaS2S.py

AlphaS2S.py

@@ -172,7 +172,7 @@ class EncoderBlock(layers.Layer):
         self.dropout2 = layers.Dropout(dropout)
     def call(self, x, mask=None, training=False):
         attn_out = self.dropout1(self.mha(x, x, x, attention_mask=mask), training=training)
-        out1 = self.norm1(x + attn_out)
+        out1 = self.norm1(attn_out)
         ffn_out = self.dropout2(self.ffn(out1), training=training)
         return self.norm2(out1 + ffn_out)
 
@@ -190,7 +190,7 @@ class DecoderBlock(layers.Layer):
         self.dropout3 = layers.Dropout(dropout)
     def call(self, x, enc_out, training=False):
         attn1 = self.dropout1(self.self_mha(x, x, x, use_causal_mask=True), training=training)
-        out1 = self.norm1(x + attn1)
+        out1 = self.norm1(attn1)
         attn2 = self.dropout2(self.cross_mha(out1, enc_out, enc_out), training=training)
         out2 = self.norm2(out1 + attn2)
         ffn_out = self.dropout3(self.ffn(out2), training=training)