Can you support the input of [Batch,samples]？

It seems that the input of multi-channel audio cannot be supported, for example, shape is [B,samples], and B refers to the number of audio.  In this case, the model needs to be processed one by one when training online, which affects the efficiency.