import torch

import numpy as np

from time import time

from math import isclose

from argparse import ArgumentParser

from utils import read, plot, calc_dist

from tsp_ip import PulpIP

from tsp_two_opt import TwoOpt

from tsp_simulated_annealing import TSPSimAnneal

from tsp_ga import TSPGA

from train import construct

def argparser():

parser = ArgumentParser()

parser.add_argument("-f", default="./ALL_tsp/ulysses16.tsp")

# PtrNet architecture

parser.add_argument('--embedding_dim', type=int, default=128, help='Embedding size')

parser.add_argument('--hidden_dim', type=int, default=512, help='Number of hidden units')

parser.add_argument('--num_lstms', type=int, default=2, help='Number of LSTM layers')

parser.add_argument('--dropout', type=float, default=0., help='Dropout value')

parser.add_argument('--bidir', default=True, action='store_true', help='Bidirectional')

parser.add_argument('--ncity', type=int, default=0)

return parser

def test1(filename):

name, ncity, D, coord = read(filename)

two_opt = TwoOpt(ncity, D)

two_opt_multi = TwoOpt(ncity, D)

simanneal = TSPSimAnneal(ncity, D)

ga = TSPGA(ncity, D)

IP = PulpIP(ncity, D, MTZ_level=2)

ts = time()

tour = two_opt.solve_two_opt(strategy="greedy_random")

total_dist = calc_dist(tour, D)

assert isclose(total_dist, two_opt.best_obj, abs_tol=1e-5)

print("\nnormal", total_dist, "\ntime", time()-ts)

ts = time()

tour = two_opt_multi.solve_multi_start_two_opt(10, strategy="greedy_random")

total_dist = calc_dist(tour, D)

assert isclose(total_dist, two_opt_multi.best_obj, abs_tol=1e-5)

print("\nmulti start", total_dist, "\ntime", time()-ts)

ts = time()

tour = simanneal.solve_simulated_annealing(T=8215972750, C=0.81, strategy="greedy_random")

total_dist = calc_dist(tour, D)

assert isclose(total_dist, simanneal.best_obj, abs_tol=1e-5)

print("\nsimulated annealing", total_dist, "\ntime", time()-ts)

ts = time()

tour = ga.solve(cxpb=0.3276646451925047, mutpb=0.6116923679473824)

total_dist = calc_dist(tour, D)

print("\nga simple", total_dist, "\ntime", time()-ts)

ts = time()

tour = IP.solve(solver_name="cplex")

total_dist = calc_dist(tour, D)

print("\nip", total_dist, "\ntime", time()-ts)

print(*tour)

#plot(tour, coord, figname=f"./{name}_{total_dist}.png")

def test2(filename):

name, ncity, D, coord = read(filename)

simanneal = TSPSimAnneal(ncity, D)

best_trial = simanneal.opt_hypara()

print("total cost", best_trial.value)

for key, val in best_trial.params.items():

print(key, val)

ga = TSPGA(ncity, D)

best_trial = ga.opt_hypara()

print("total cost", best_trial.value)

for key, val in best_trial.params.items():

print(key, val)

def test3(params):

name, ncity, D, coord = read(params.f)

params.ncity = ncity

model, _, _ = construct("PtrNet", params, is_train=False)

_input = torch.Tensor([np.asarray([x for x in coord.values()])])

ts = time()

_, tour = model(_input)

tour = list(tour.detach().numpy()[0])

total_dist = calc_dist(tour, D)

print("\nPtrNet", total_dist, "\ntime", time()-ts)

print(*tour)

model, _, _ = construct("NeuralCombOptRL", params, is_train=False)

_input = torch.Tensor([np.asarray([x for x in coord.values()])])

ts = time()

_, _, _, tour = model(_input)

tour = list(tour.detach().numpy()[0])

total_dist = calc_dist(tour, D)

print("\nNeuralCombOptRL", total_dist, "\ntime", time()-ts)

print(*tour)

if __name__=="__main__":

#import pyscipopt

parser = argparser()

args = parser.parse_args()

test1(args.f)

# test2(args.f)

test3(args)

# ToDo

"""

+ optunaによるパラメータチューニングの実装

+ amplifyによるQAの実装

+ MTZ制約を強くした版の実装

+ 凸包による初期化

+ Tabu search

+ ant colony

+ swapをgreedyにやる

+ beam search みたいな

+ GAのバリエーション

+ multi start parallel

"""