Update AlphaS2S.py

AlphaS2S.py

@@ -255,7 +255,7 @@ def create_lr_schedule(initial_lr=5e-5, decay_steps=10000, decay_rate=0.9):
 
 with strategy.scope():
     # ⚠️ 수정: chat_vocab_size 대신 정의된 vocab_size 사용
-    chat_model = Transformer(num_layers=2, d_model=384, num_heads=6, dff=1024, input_vocab_size=vocab_size, target_vocab_size=vocab_size, max_len=256, dropout=0.1)
+    chat_model = Transformer(num_layers=2, d_model=320, num_heads=5, dff=1024, input_vocab_size=vocab_size, target_vocab_size=vocab_size, max_len=256, dropout=0.1)
     
     dummy_input = {
         "enc_inputs": tf.zeros((1, max_len), dtype=tf.int32),