#include <iostream>
#include <stdio.h>
#include <cmath>
#include <vector>
#include <tuple>
#include <bitset>
#include <map>
#include <queue>
#include <time.h>
#include <utility>
#include <numeric>
#include <deque>
#include <cassert>
#include <sstream>
#include <cctype>
#include <unordered_map>
#include <algorithm>
#include <chrono>
#define ll long long
#include <time.h>
#define mod % 1000000007
#define rep(i,n) for(long long i = 0;i < n;i ++)
ll MAX = 5'000'000'000'000'000'000LL;
using namespace std;

ll A = 17;
ll A2 = 1LL<<17;

struct fenwick_tree{
  vector<vector<ll>> v;
  void start(){
    vector<ll> V(A2,0);
    rep(i,A)v.push_back(V);
  }
  void add(ll n,ll val){
    for(int i = 0;i < A;i ++){
      v[i][n>>i] += val;
    }
  }
  ll sum(ll n){
      n ++;
      ll s = 0;
      ll u = 0;
      ll d = A2;
      for(ll i = A-1;i >= 0;i --){
          ll m = (u+d)/2;
          if(m == n){
              s += v[i][u>>i];
              break;
          }
          if(m<n){
              s += v[i][u>>i];
              u = m;
          }
          if(m>n){
              d = m;
          }
      }
      return s;
  }
};

vector<ll> conv(vector<ll> a){
  auto b = a;
  sort(b.begin(),b.end());
  ll c = b.size();
  map<ll,ll> mp;
  for(int i = c-2;i >= 0;i --){
    if(b[i] == b[i+1])b.erase(b.begin()+i);
  }
  c = b.size();
  rep(i,c){
    mp[b[i]] = i+2;
  }
  rep(i,int(a.size())){
    a[i] = mp[a[i]];
  }
  return a;
}

int main(){
  vector<ll> exp;
  exp.push_back(1);
  rep(i,1000000)exp.push_back((exp[i]*2)mod);
  ll n;
  cin >> n;
  vector<ll> a(n);
  rep(i,n)cin >> a[i];
  a = conv(a);
  vector<ll> lup(n);
  vector<ll> ldown(n);
  fenwick_tree f;
  f.start();
  rep(i,n){
    lup[i] = (f.sum(A2-2)-f.sum(a[i])) mod;
    ldown[i] = f.sum(a[i]-1) mod;
    f.add(a[i],exp[i]);
  }
  vector<ll> rup(n);
  vector<ll> rdown(n);
  fenwick_tree g;
  g.start();
  rep(i,n){
    rup[n-1-i] = (g.sum(A2-2)-g.sum(a[n-1-i])) mod;
    rdown[n-1-i] = g.sum(a[n-1-i]-1) mod;
    g.add(a[n-1-i],exp[i]);
  }
  ll ans = 0;
  rep(i,n){
      ans += (rup[i]*lup[i]) mod;
      ans += (ldown[i]*rdown[i]) mod;
  }
  cout << ans mod << endl;
  
  
  return 0;
}