mx.sym.batch_dot calculate matrix multiplication

[Aruen24]

I use the mx.sym.batch_dot to calculate matrix multiplication, but my network report error:
ValueError: Input contains NaN, infinity or a value too large for dtype('float64').
why???