NEAT - population number varies every generation - machine-learning

Question

I'm trying to implement my own neat implementation and I can't get myself to understand how speciation works
I tried my best to follow the pesudocode I found in this paper (start of page 13)
but I'm think I'm doing it really wrong but I don't understand the right way to do it, here is my code
the sepciate function that splits the population into species:
function speciate(population, species=[]) {
let newSpecies = [...species];
for(const net of population) {
let placed = false;
for(const s of newSpecies) {
for(const member of s) {
if(sh(net, member)) {
s.push(net);
placed = true;
break;
}
}
if(placed) break;
}
if(!placed) {
newSpecies.push([net]);
}
}
return newSpecies;
}
the repopulation function that generates a new population using the number of offsprings:
function repopulate(popCount, species) {
let globalAvg = 0;
species.forEach(s => {
globalAvg += s.reduce((P, net) => P + net.genome.fitness, 0) / s.length;
});
let newPop = [];
for(const s of species) {
let N = popCount;
let sAvg = s.reduce((P, net) => P + net.genome.fitness, 0) / s.length;
let offspringCount = (sAvg / globalAvg) * N;
for(let i = 0; i < offspringCount; i++) {
let parent1 = wheelSelect(s);
let parent2 = wheelSelect(s);
let child = parent1.genome.crossover(parent2.genome);
child.mutateAddNeuron(0.01);
child.mutateAddConnection(0.01);
child.mutateWeight(0.01);
child.mutateEnabledToggle(0.01);
child.layerNeurons();
let net = new NeuralNetwork();
net.wireUsingGenome(child);
newPop.push(net);
}
}
return newPop;
}
the problem I'm facing is that the population number seems to change every new generation sometimes it goes up and sometimes down, so I'm gussing I'm calculating the offspring count wrong or my speciation isn't working correctly but I can't figure it out
any help is appreciated!