I was making a neural network program, single thread, when I noticed the intermediate steps were coming out too large with sigmoid activation. Forward propa
