play_ground.py

import numpy as np
from numpy import genfromtxt
import torch
from tqdm import tqdm


problem_type = 'JSSP'
j = 30
m = 20
data = np.load('./validation_data/{}_validation_data_and_Cmax_{}x{}_[1,99].npy'.format(problem_type, j, m))
print(data[0, 1, 0])
# data[:, 1, :, :] += 1
# print(data[0, 1, 0])
# np.save('./validation_data/validation_data_and_Cmax_{}x{}_[1,99].npy'.format(j, m), data)

j = 15
m = 15
tp = 'tai'
data = np.load('./test_data_jssp/{}{}x{}.npy'.format(tp, j, m))
result = np.load('./test_data_jssp/{}{}x{}_result.npy'.format(tp, j, m))
# time = np.load('./test_data_jssp/{}{}x{}_time.npy'.format(tp, j, m))
print(data[0].shape)
print(result)
# print(time)


# j = 20
# m = 5
# start_segment_flag = 0
# end_segment_flag = 10
# tp = 'tai'
# result = np.load('./test_data_fssp/ortools_result_FSSP-{}{}x{}[{},{}]_result.npy'.format(tp, j, m, start_segment_flag, end_segment_flag))
# time = np.load('./test_data_fssp/ortools_result_FSSP-{}{}x{}[{},{}]_time.npy'.format(tp, j, m, start_segment_flag, end_segment_flag))
# print(result)
# print(time)


j = 20
m = 10
horizon_l2s = 3
horizon_tb = 1
l2s_result = np.load('l2s_result_{}x{}.npy'.format(j, m))
ts_result = np.load('tabu_search_result_{}x{}.npy'.format(j, m))
opt_result = np.load('./test_data_jssp/syn{}x{}_result.npy'.format(j, m))
# print(l2s_result.shape)
# print(ts_result.shape)
# print(l2s_result)
# print(ts_result)

relative_error = (l2s_result[horizon_l2s] - ts_result[horizon_tb]) / ts_result[horizon_tb]
print('relative gap to tb:', relative_error.mean())
print('outperform ratio:', (relative_error < 0).sum() / relative_error.shape[0])

gap_l2s = (l2s_result[horizon_l2s] - opt_result) / opt_result
gap_tb = (ts_result[horizon_tb] - opt_result) / opt_result
print('opt gap of l2s:', gap_l2s.mean())
print('opt gap of tb:', gap_tb.mean())
print('improvement in terms of opt gap compared with tabu:', ((gap_l2s.mean() - gap_tb.mean())/gap_tb.mean()))