forked from pengguo318/FJSPDRL
-
Notifications
You must be signed in to change notification settings - Fork 0
/
validation.py
175 lines (148 loc) · 8.36 KB
/
validation.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
from epsGreedyForMch import PredictMch
from mb_agg import *
from Params import configs
from copy import deepcopy
from FJSP_Env import FJSP,DFJSP_GANTT_CHART
from mb_agg import g_pool_cal
import copy
from agent_utils import sample_select_action
from agent_utils import greedy_select_action
import numpy as np
import torch
import matplotlib.pyplot as plt
from Params import configs
def validate(vali_set,batch_size, policy_jo,policy_mc):
policy_job = copy.deepcopy(policy_jo)
policy_mch = copy.deepcopy(policy_mc)
policy_job.eval()
policy_mch.eval()
def eval_model_bat(bat,i):
C_max = []
with torch.no_grad():
data = bat.numpy()
env = FJSP(n_j=configs.n_j, n_m=configs.n_m)
gantt_chart = DFJSP_GANTT_CHART( configs.n_j, configs.n_m)
device = torch.device(configs.device)
g_pool_step = g_pool_cal(graph_pool_type=configs.graph_pool_type,
batch_size=torch.Size(
[batch_size, configs.n_j * configs.n_m, configs.n_j * configs.n_m]),
n_nodes=configs.n_j * configs.n_m,
device=device)
adj, fea, candidate, mask, mask_mch, dur, mch_time, job_time = env.reset(data)
j = 0
ep_rewards = - env.initQuality
rewards = []
env_mask_mch = torch.from_numpy(np.copy(mask_mch)).to(device)
env_dur = torch.from_numpy(np.copy(dur)).float().to(device)
pool=None
while True:
env_adj = aggr_obs(deepcopy(adj).to(device).to_sparse(), configs.n_j * configs.n_m)
env_fea = torch.from_numpy(np.copy(fea)).float().to(device)
env_fea = deepcopy(env_fea).reshape(-1, env_fea.size(-1))
env_candidate = torch.from_numpy(np.copy(candidate)).long().to(device)
env_mask = torch.from_numpy(np.copy(mask)).to(device)
env_mch_time = torch.from_numpy(np.copy(mch_time)).float().to(device)
# env_job_time = torch.from_numpy(np.copy(job_time)).float().to(device)
action, a_idx, log_a, action_node, _, mask_mch_action, hx = policy_job(x=env_fea,
graph_pool=g_pool_step,
padded_nei=None,
adj=env_adj,
candidate=env_candidate
, mask=env_mask
, mask_mch=env_mask_mch
, dur=env_dur
, a_index=0
, old_action=0
,mch_pool=pool
,old_policy=True,
T=1
,greedy=True
)
pi_mch,pool = policy_mch(action_node, hx, mask_mch_action, env_mch_time)
_, mch_a = pi_mch.squeeze(-1).max(1)
adj, fea, reward, done, candidate, mask,job,_,mch_time,job_time = env.step(action.cpu().numpy(), mch_a,gantt_chart)
#rewards += reward
j += 1
if env.done():
plt.savefig("./3020_%s.svg"%i, format='svg',dpi=300, bbox_inches='tight')
#plt.show()
break
cost = env.mchsEndTimes.max(-1).max(-1)
C_max.append(cost)
return torch.tensor(cost)
#make_spans.append(rewards - env.posRewards)
#print(env.mchsStartTimes,env.mchsEndTimes,env.opIDsOnMchs)
#print('REWARD',rewards - env.posRewards)
totall_cost = torch.cat([eval_model_bat(bat,i) for i,bat in enumerate(vali_set)], 0)
return totall_cost
if __name__ == '__main__':
from uniform_instance import uni_instance_gen,FJSPDataset
import numpy as np
import time
import argparse
from Params import configs
parser = argparse.ArgumentParser(description='Arguments for ppo_jssp')
parser.add_argument('--Pn_j', type=int, default=30, help='Number of jobs of instances to test')
parser.add_argument('--Pn_m', type=int, default=20, help='Number of machines instances to test')
parser.add_argument('--Nn_j', type=int, default=30, help='Number of jobs on which to be loaded net are trained')
parser.add_argument('--Nn_m', type=int, default=20, help='Number of machines on which to be loaded net are trained')
parser.add_argument('--low', type=int, default=-99, help='LB of duration')
parser.add_argument('--high', type=int, default=99, help='UB of duration')
parser.add_argument('--seed', type=int, default=200, help='Cap seed for validate set generation')
parser.add_argument('--n_vali', type=int, default=100, help='validation set size')
params = parser.parse_args()
N_JOBS_P = params.Pn_j
N_MACHINES_P = params.Pn_m
LOW = params.low
HIGH = params.high
N_JOBS_N = params.Nn_j
N_MACHINES_N = params.Nn_m
from torch.utils.data import DataLoader
from PPOwithValue import PPO
import torch
import os
from torch.utils.data import Dataset
ppo = PPO(configs.lr, configs.gamma, configs.k_epochs, configs.eps_clip,
n_j=N_JOBS_P,
n_m=N_MACHINES_P,
num_layers=configs.num_layers,
neighbor_pooling_type=configs.neighbor_pooling_type,
input_dim=configs.input_dim,
hidden_dim=configs.hidden_dim,
num_mlp_layers_feature_extract=configs.num_mlp_layers_feature_extract,
num_mlp_layers_actor=configs.num_mlp_layers_actor,
hidden_dim_actor=configs.hidden_dim_actor,
num_mlp_layers_critic=configs.num_mlp_layers_critic,
hidden_dim_critic=configs.hidden_dim_critic)
filepath = 'saved_network'
filepath = os.path.join(filepath, 'FJSP_J%sM%s' % (30,configs.n_m))
#filepath = os.path.join(filepath, '%s_%s' % (0,239))
filepath = os.path.join(filepath, 'best_value0')
job_path = './{}.pth'.format('policy_job')
mch_path = './{}.pth'.format('policy_mch')
'''filepath = 'saved_network'
filepath = os.path.join(filepath,'%s'%19)
job_path = './{}.pth'.format('policy_job'+str(N_JOBS_N) + '_' + str(N_MACHINES_N) + '_' + str(LOW) + '_' + str(HIGH))
mch_path = './{}.pth'.format('policy_mch'+ str(N_JOBS_N) + '_' + str(N_MACHINES_N) + '_' + str(LOW) + '_' + str(HIGH))'''
job_path = os.path.join(filepath,job_path)
mch_path = os.path.join(filepath, mch_path)
ppo.policy_job.load_state_dict(torch.load(job_path))
ppo.policy_mch.load_state_dict(torch.load(mch_path))
num_val = 10
batch_size = 1
SEEDs = [200]
result = []
loade = False
for SEED in SEEDs:
mean_makespan = []
#np.random.seed(SEED)
if loade:
validat_dataset = np.load(file="FJSP_J%sM%s_unew_test_data.npy" % (configs.n_j, configs.n_m))
print(validat_dataset.shape[0])
else:
validat_dataset = FJSPDataset(configs.n_j, configs.n_m, configs.low, configs.high, num_val, SEED)
valid_loader = DataLoader(validat_dataset, batch_size=batch_size)
vali_result = validate(valid_loader,batch_size, ppo.policy_job, ppo.policy_mch)
#mean_makespan.append(vali_result)
print(vali_result,np.array(vali_result).mean())
# print(min(result))