#pragma GCC target("avx2")
#pragma GCC optimize("O3")
#pragma GCC optimize("unroll-loops")
#include <atcoder/all>
#include <iostream> // cout, endl, cin
#include <string> // string, to_string, stoi
#include <vector> // vector
#include <algorithm> // min, max, swap, sort, reverse, lower_bound, upper_bound
#include <utility> // pair, make_pair
#include <tuple> // tuple, make_tuple
#include <cstdint> // int64_t, int*_t
#include <cstdio> // printf
#include <map> // map
#include <queue> // queue, priority_queue
#include <set> // set
#include <stack> // stack
#include <deque> // deque
#include <unordered_map> // unordered_map
#include <unordered_set> // unordered_set
#include <bitset> // bitset
#include <cctype> // isupper, islower, isdigit, toupper, tolower
#include <math.h>
#include <iomanip>
#include <functional>
using namespace std;  
using namespace atcoder;
#define rep(i, n) for (int i = 0; i < (int)(n); i++)
#define repi(i, a, b) for (int i = (int)(a); i < (int)(b); i++)
typedef long long ll;
typedef unsigned long long ull;
const ll inf=1e18;  
using graph = vector<vector<int> > ;
using P= pair<ll,ll>;  
using vi=vector<int>;
using vvi=vector<vi>;
using vll=vector<ll>; 
using vvll=vector<vll>;
using vp=vector<P>;
using vvp=vector<vp>;
using vd=vector<double>;
using vvd =vector<vd>;
//string T="ABCDEFGHIJKLMNOPQRSTUVWXYZ";
//string S="abcdefghijklmnopqrstuvwxyz";
//g++ main.cpp -std=c++17 -I .  
//cout <<setprecision(20);
//cout << fixed << setprecision(10);
//cin.tie(0); ios::sync_with_stdio(false);
const double PI = acos(-1);
int vx[]={0,1,0,-1,-1,1,1,-1},vy[]={1,0,-1,0,1,1,-1,-1};
void putsYes(bool f){cout << (f?"Yes":"No") << endl;}
void putsYES(bool f){cout << (f?"YES":"NO") << endl;}
void putsFirst(bool f){cout << (f?"First":"Second") << endl;}
void debug(int test){cout << "TEST" << " " << test << endl;}
ll pow_pow(ll x,ll n,ll mod){
    if(n==0) return 1; 
    x%=mod;
    ll res=pow_pow(x*x%mod,n/2,mod);
    if(n&1)res=res*x%mod;
    return res;
}
ll gcd(ll x,ll y){
    if(y==0)return x;
    return gcd(y,x%y);
}
 
ll lcm(ll x,ll y){
    return ll(x/gcd(x,y))*y;
}
template<class T> bool chmin(T& a, T b) {
    if (a > b) {
        a = b;
        return true;
    }
    else return false;
}
template<class T> bool chmax(T& a, T b) {
    if (a < b) {
        a = b;
        return true;
    }
    else return false;
}
 
// https://youtu.be/L8grWxBlIZ4?t=9858
// https://youtu.be/ERZuLAxZffQ?t=4807 : optimize
// https://youtu.be/8uowVvQ_-Mo?t=1329 : division
ll mod =998244353;
struct mint {
  ll x; // typedef long long ll;
  mint(ll x=0):x((x%mod+mod)%mod){}
  mint operator-() const { return mint(-x);}
  mint& operator+=(const mint a) {
    if ((x += a.x) >= mod) x -= mod;
    return *this;
  }
  mint& operator-=(const mint a) {
    if ((x += mod-a.x) >= mod) x -= mod;
    return *this;
  }
  mint& operator*=(const mint a) { (x *= a.x) %= mod; return *this;}
  mint operator+(const mint a) const { return mint(*this) += a;}
  mint operator-(const mint a) const { return mint(*this) -= a;}
  mint operator*(const mint a) const { return mint(*this) *= a;}
  mint pow(ll t) const {
    if (!t) return 1;
    mint a = pow(t>>1);
    a *= a;
    if (t&1) a *= *this;
    return a;
  }
 
  // for prime mod
  mint inv() const { return pow(mod-2);}
  mint& operator/=(const mint a) { return *this *= a.inv();}
  mint operator/(const mint a) const { return mint(*this) /= a;}
};
istream& operator>>(istream& is, const mint& a) { return is >> a.x;}
ostream& operator<<(ostream& os, const mint& a) { return os << a.x;}
// combination mod prime
// https://www.youtube.com/watch?v=8uowVvQ_-Mo&feature=youtu.be&t=1619
struct combination {
  vector<mint> fact, ifact;
  combination(int n):fact(n+1),ifact(n+1) {
    //assert(n < mod);
    fact[0] = 1;
    for (int i = 1; i <= n; ++i) fact[i] = fact[i-1]*i;
    ifact[n] = fact[n].inv();
    for (int i = n; i >= 1; --i) ifact[i-1] = ifact[i]*i;
  }
  mint operator()(int n, int k) {
    if (k < 0 || k > n) return 0;
    return fact[n]*ifact[k]*ifact[n-k];
  } mint p(int n, int k) { return fact[n]*ifact[n-k]; } } c(500000); 
  using vm=vector<mint> ;
using vvm=vector<vm> ;
 
 
ll sqrt_(ll x) {
  ll l = 0, r = ll(3e9)+1;
  while (l+1<r) {
    ll c = (l+r)/2;
    if (c*c <= x) l = c; else r = c;
  }
  return l;
}
 
 
int valid(int x,int y,int h,int w){
  if(x>=0 && y>=0 && x<h && y<w)return 1;
  else return 0;
}
 
struct edge{
  int to; ll cost;
  edge(int to,ll cost) : to(to),cost(cost){}
};
// https://youtu.be/ylWYSurx10A?t=2352
template<typename T>
struct Matrix {
  int h, w;
  vector<vector<T>> d;
  Matrix() {}
  Matrix(int h, int w, T val=0): h(h), w(w), d(h, vector<T>(w,val)) {}
  Matrix& unit() {
   // assert(h == w);
    rep(i,h) d[i][i] = 1;
    return *this;
  }
  const vector<T>& operator[](int i) const { return d[i];}
  vector<T>& operator[](int i) { return d[i];}
  Matrix operator*(const Matrix& a) const {
 //   assert(w == a.h);
    Matrix r(h, a.w);
    rep(i,h)rep(k,w)rep(j,a.w) {
      r[i][j] += d[i][k]*a[k][j];
    }
    return r;
  }
  Matrix pow(long long t) const {
  //  assert(h == w);
    if (!t) return Matrix(h,h).unit();
    if (t == 1) return *this;
    Matrix r = pow(t>>1);
    r = r*r;
    if (t&1) r = r*(*this);
    return r;
  }
  // https://youtu.be/-j02o6__jgs?t=11273
  /* mint only
  mint det() {
    assert(h == w);
    mint res = 1;
    rep(k,h) {
      for (int i = k; i < h; ++i) {
        if (d[i][k] == 0) continue;
        if (i != k) {
          swap(d[i],d[k]);
          res = -res;
        }
      }
      if (d[k][k] == 0) return 0;
      res *= d[k][k];
      mint inv = mint(1)/d[k][k];
      rep(j,h) d[k][j] *= inv;
      for (int i = k+1; i < h; ++i) {
        mint c = d[i][k];
        for (int j = k; j < h; ++j) d[i][j] -= d[k][j]*c;
      }
    }
    return res;
  }
  //*/
};
using ve=vector<edge>;
using vve=vector<ve>;
void solve(int test){
  int m,n; cin >> m >> n;
  vi a(n);rep(i,n)cin >> a[i];
  vi cnt(m+1);
  rep(i,n)cnt[a[i]]++;
  vp p;
  {
    vi v;
    for(int i=1; i<=m; i++)if(cnt[i])v.push_back(cnt[i]);
    vi cnt2(n+1);
    for(auto u:v)cnt2[u]++;
    for(int i=1; i<=n; i++)if(cnt2[i])p.push_back(P(i,cnt2[i]));

  }
  vm kai(n+1,1);for(int i=1; i<=n; i++)kai[i]=kai[i-1]*i;
  vm kaiinv(n+1,1);for(int i=1; i<=n; i++)kaiinv[i]=kai[i].inv();
  vll g(n+1,1);
  vll g2(n+1,1);
  mint now=1;
  mint now2=1;
  for(int i=0; i<=n; i++){
    now=1;
    for(auto u:p){
      int num=u.first;
      now2=c(i,num)*kai[num];
      now2=pow_pow(now2.x,u.second,mod);
      now*=now2;
    }
    now*=kaiinv[i];
    g[i]=now.x;
  }
  mint base=1;
  for(int i=0; i<=n; i++){
    now=base;
    now*=kaiinv[i];
    g2[i]=now.x;
    base*=(mod-1);
  }
  auto conv=convolution(g,g2);
  mint ans=0;
  for(int i=1; i<=n; i++){
    ans+=kai[i]*conv[i];
  }
  cout << ans << endl;
}
//g++ main.cpp -std=c++17 -I .
int main(){cin.tie(0);ios::sync_with_stdio(false);
  int t=1; //cin >> t;
  rep(test,t)solve(test);
}