-
Notifications
You must be signed in to change notification settings - Fork 7
/
Copy pathdriver.py
234 lines (197 loc) · 10.6 KB
/
driver.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
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
import ray
import torch
import torch.optim as optim
from torch.utils.tensorboard import SummaryWriter
import os
import copy
import numpy as np
import time
from scipy.stats import ttest_rel
from env import Env
from model import Model
from config import config
from runner import Runner
from runner_for_test import RayTestRunner
ray.init()
cfg = config()
writer = SummaryWriter(cfg.train_path)
if not os.path.exists(cfg.model_path):
os.makedirs(cfg.model_path)
if not os.path.exists(cfg.gifs_path):
os.makedirs(cfg.gifs_path)
def main(cfg):
device = cfg.device
global_model = Model(cfg)
global_model.share_memory()
global_model.to(device)
optimizer = optim.AdamW(global_model.parameters(), lr=cfg.lr)
lr_decay = optim.lr_scheduler.StepLR(optimizer, step_size=256, gamma=0.96)
meta_agent_list = [Runner.remote(metaAgentID=i, cfg=cfg) for i in range(cfg.meta_agent_amount)]
# info for tensorboard
average_loss = 0
average_advantage = 0
average_grad_norm = 0
average_rewards = 0
average_max_length = 0
average_entropy = 0
global_step = 0
if cfg.load_model:
checkpoint = torch.load(cfg.model_path + '/model_states.pth')
global_step = checkpoint['step']
global_model.load_state_dict(checkpoint['model'])
optimizer.load_state_dict(checkpoint['optimizer'])
lr_decay.load_state_dict(checkpoint['lr_decay'])
print("load model at", global_step)
print(optimizer.state_dict()['param_groups'][0]['lr'])
# get global network weights
global_weights = global_model.state_dict()
# update local network
update_local_network_job_list = []
for i, meta_agent in enumerate(meta_agent_list):
update_local_network_job_list.append(meta_agent.set_model_weights.remote(global_weights))
baseline_weights = copy.deepcopy(global_weights)
update_baseline_network_job_list = []
for i, meta_agent in enumerate(meta_agent_list):
update_baseline_network_job_list.append(meta_agent.set_baseline_model_weights.remote(baseline_weights))
baseline_value = None
test_set = np.random.randint(low=0, high=1e8, size=[2048 // cfg.meta_agent_amount, cfg.meta_agent_amount])
try:
while True:
global_step += 1
#print(global_step)
sample_job_list = []
for i, meta_agent in enumerate(meta_agent_list):
sample_job_list.append(meta_agent.sample.remote())
if global_step % cfg.batch_size == 0:
# get gradient and loss from runner
get_gradient_job_list = []
for i, meta_agent in enumerate(meta_agent_list):
get_gradient_job_list.append(meta_agent.return_gradient.remote())
gradient_set_id, _ = ray.wait(get_gradient_job_list, num_returns=cfg.meta_agent_amount)
gradient_loss_set = ray.get(gradient_set_id)
for gradients, loss, grad_norm, advantage, max_length,entropy,reward in gradient_loss_set:
average_max_length += max_length
average_loss += loss
average_advantage += advantage
average_grad_norm += grad_norm
average_entropy += entropy
average_rewards += reward
optimizer.zero_grad()
for g, global_param in zip(gradients, global_model.parameters()):
global_param._grad = g
# update networks
optimizer.step()
lr_decay.step()
update_local_network_job_list = []
for i, meta_agent in enumerate(meta_agent_list):
update_local_network_job_list.append(meta_agent.set_model_weights.remote(global_weights))
# tensorboard update
if global_step % cfg.tensorboard_batch == 0:
writer.add_scalar('loss/loss',
average_loss / (cfg.meta_agent_amount * cfg.tensorboard_batch / cfg.batch_size),
global_step)
average_loss = 0
writer.add_scalar('loss/entropy',
average_entropy / (cfg.meta_agent_amount * cfg.tensorboard_batch / cfg.batch_size),
global_step)
average_entropy = 0
writer.add_scalar('loss/advantage',
average_advantage / (
cfg.meta_agent_amount * cfg.tensorboard_batch / cfg.batch_size),
global_step)
average_advantage = 0
writer.add_scalar('grad/grad_norm',
average_grad_norm / (
cfg.meta_agent_amount * cfg.tensorboard_batch / cfg.batch_size),
global_step)
average_grad_norm = 0
writer.add_scalar('perf/reward', average_rewards / (cfg.meta_agent_amount * cfg.tensorboard_batch / cfg.batch_size), global_step)
average_rewards = 0
writer.add_scalar('perf/max_length', average_max_length / (
cfg.meta_agent_amount * cfg.tensorboard_batch / cfg.batch_size), global_step)
average_max_length = 0
# save model
if global_step % cfg.log_size == 0:
model_states = {"model": global_model.state_dict(),
"optimizer": optimizer.state_dict(),
"lr_decay": lr_decay.state_dict(),
"step": global_step}
torch.save(obj=model_states, f=cfg.model_path + '/model_states.pth')
# update baseline model every 1024 steps
if global_step % (2048) == 0:
# stop the training
ray.wait(update_local_network_job_list, num_returns=cfg.meta_agent_amount)
for a in meta_agent_list:
ray.kill(a)
torch.cuda.empty_cache()
time.sleep(5)
print('evaluate baseline model at ', global_step)
# test the baseline model on the new test set
if baseline_value is None:
test_agent_list = [RayTestRunner.remote(metaAgentID=i, cfg=cfg, decode_type='greedy') for i in
range(cfg.meta_agent_amount)]
update_local_network_job_list = []
for _, test_agent in enumerate(test_agent_list):
update_local_network_job_list.append(test_agent.set_weights.remote(baseline_weights))
max_length_list = []
for i in range(2048 // cfg.meta_agent_amount):
sample_job_list = []
for j, test_agent in enumerate(test_agent_list):
env = Env(cfg, test_set[i][j])
sample_job_list.append(test_agent.sample.remote(env))
sample_done_id, _ = ray.wait(sample_job_list, num_returns=cfg.meta_agent_amount)
rewards = ray.get(sample_done_id)
max_length_list = max_length_list + rewards
baseline_value = torch.stack(max_length_list).squeeze(0).cpu().numpy()
for a in test_agent_list:
ray.kill(a)
# test the current model's performance
test_agent_list = [RayTestRunner.remote(metaAgentID=i, cfg=cfg, decode_type='greedy') for i in
range(cfg.meta_agent_amount)]
update_local_network_job_list = []
for _, test_agent in enumerate(test_agent_list):
update_local_network_job_list.append(test_agent.set_weights.remote(global_weights))
max_length_list = []
for i in range(2048 // cfg.meta_agent_amount):
sample_job_list = []
for j, test_agent in enumerate(test_agent_list):
env = Env(cfg, test_set[i][j])
sample_job_list.append(test_agent.sample.remote(env))
sample_done_id, _ = ray.wait(sample_job_list, num_returns=cfg.meta_agent_amount)
rewards = ray.get(sample_done_id)
max_length_list = max_length_list + rewards
test_value = torch.stack(max_length_list).squeeze(0).cpu().numpy()
# restart training
print('lr', optimizer.state_dict()['param_groups'][0]['lr'])
for a in test_agent_list:
ray.kill(a)
time.sleep(5)
meta_agent_list = [Runner.remote(metaAgentID=i, cfg=cfg) for i in range(cfg.meta_agent_amount)]
for i, meta_agent in enumerate(meta_agent_list):
update_local_network_job_list.append(meta_agent.set_model_weights.remote(global_weights))
update_baseline_network_job_list = []
for i, meta_agent in enumerate(meta_agent_list):
update_baseline_network_job_list.append(meta_agent.set_baseline_model_weights.remote(baseline_weights))
# update baseline if the model improved more than 5%
print('test value', test_value.mean())
print('baseline value', baseline_value.mean())
if test_value.mean() < baseline_value.mean():
_, p = ttest_rel(test_value, baseline_value)
print('p value', p)
if p < 0.05:
print('update baseline model at ', global_step)
global_weights = global_model.state_dict()
baseline_weights = copy.deepcopy(global_weights)
update_baseline_network_job_list = []
for i, meta_agent in enumerate(meta_agent_list):
update_baseline_network_job_list.append(meta_agent.set_baseline_model_weights.remote(baseline_weights))
test_set = np.random.randint(low=0, high=1e8,
size=[2048 // cfg.meta_agent_amount, cfg.meta_agent_amount])
print('update test set')
baseline_value = None
except KeyboardInterrupt:
print("CTRL-C pressed. killing remote workers")
for a in meta_agent_list:
ray.kill(a)
if __name__ == '__main__':
main(cfg)