Add attention_bias
argument in transformer block and transformer layer modules, addressing change in MCore
#6344
Loading