LSTM input in nn.LSTM

I have a multivariate time series with length 48 and at each time step I have 42 variables so for each timestep I get a matix (42 * 30) where 30 is number of times that variable changed during one time step.
and using “ALL THOSE” 42 variables at 48 time steps (x=48 * 42*30)–(y=1/0) they correspond to a y (it either takes a 1 or a 0).

Now I have 4000 such x and corresponding 4000 such y’s.
I am not able model the input to the Lstm.
I went through many tutorials and online forums but couldnt find anthing.
Can anyone please help me with this.