Hello, I have a matmul core which I use for computing dense layers. For batch_size=1 it works fine, but for larger batches, all except batch one yield wrong results. I figure this has to do with the bias, my kernel expects a bias of size (batch_size, out_dim) but a dense layer has one of shape (out_dim). Is there a way to replicate the bias along the axis? I found the topi expand_like function, which seems to introduce an error, and is not resolved at compile time. Is there something that statically expands the dimension of the bias and just replicates the value during compilation?