// needs to be accounted for in the trainer
std::max(0ll, std::min(127ll, (((long long)input[i] * input[i]) >> (2 * WeightScaleBits)) / 128)));
}
// needs to be accounted for in the trainer
std::max(0ll, std::min(127ll, (((long long)input[i] * input[i]) >> (2 * WeightScaleBits)) / 128)));
}