Reading MNIST Handwritten digit data in system Verilog - deep-learning

Question

My goal is to read data from MNIST file using system Verilog and perform a dot product between two image shape numbers by reading these files in a test vector arrays. And display that result. noting that, I have taken only the first 200X785 mnist file data and converted to a txt file so I could use $readmemb in system verilog even though the data inside these files are in gray scale 0-255 can you guys help?
`begin_keywords "1800-2012" //Use keyword list from SystemVerilog-2012.
module read_write ( // This is DUT. It specifies input signals to drive the DUT.
// input from MNIST files
input logic [784:0] A1, A2,output logic [784:0] product_0_1,);
timeunit 1ns/1ns; // Simulation time unit of 1 ns, with precision of 1 ns.
// at start of test load testvectors
initial
begin
$readmemb ("D:/New Folder/MNIST200.txt",testvectors);//the mnist is stored in a test
vector file "MNIST200.txt".
//using a for loop to read in each of the values within the test vector array
for (int i=0;i<785; i=i+1)
begin
$display("%b", testvectors[i]);
end
$finish;
end
always_comb
begin
//Sum of product
// to identify the rectified linear unit. activation function
product_0_1 <= sum (A1 * A2);
end
endmodule:read_write;
`end_keywords