-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtrain_sub_network.py
158 lines (126 loc) · 6.47 KB
/
train_sub_network.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
import os
import sys
import time
import torch
import logging
import dgl
import torch.utils
import psutil
import torch.nn.functional as F
import torch.backends.cudnn as cudnn
from models.search_space.network_f2 import SubNetwork
from models.component_prediction.gat_predictor import GatPredictor
from models.component_prediction.gcn_predictor import GcnPredictor
from util.logging import init_logger
from util.file_handler import deserialize
from models.component_prediction.prediction_dataset import PredictionDataset, FakeDataset
from models.component_prediction.hyperparameter_configuration import hyperparameters
torch.set_printoptions(precision=4)
import pickle
import numpy as np
from util.utils import count_parameters_in_MB
from util.parser import get_args
from nas_trainer.subnet_trainer import SubPredictionTrainer, SubFakePredictionTrainer
from nas_trainer.prediction_evaluator import PredictionEvaluator, PredictionCsEvaluator
def get_dataset(data_filename, dataset_path, transform_to_emb_size, mode='train'):
if os.path.exists(f'{dataset_path}_{mode}'):
with open(f'{dataset_path}_{mode}', 'rb') as f:
dataset = pickle.load(f)
else:
dataset = PredictionDataset(data_filename, mode, transform_to_emb_size)
with open(f'{dataset_path}_{mode}', 'wb') as f:
pickle.dump(dataset, f)
return dataset
def get_test_dataset(data_filename, dataset_path, transform_to_emb_size, mode='train'):
if os.path.exists(f'{dataset_path}_{mode}_test'):
with open(f'{dataset_path}_{mode}_test', 'rb') as f:
dataset = pickle.load(f)
else:
dataset = PredictionDataset(data_filename, mode, transform_to_emb_size)
dataset.set_len()
dataset.graphs = dataset.graphs[:100]
dataset.labels = dataset.labels[:100]
with open(f'{dataset_path}_{mode}_test', 'wb') as f:
pickle.dump(dataset, f)
dataset.set_len()
return dataset
def set_seed(seed):
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
dgl.seed(seed)
np.random.seed(seed)
def main(args):
global device
device = torch.device('cuda:%d' % args.gpu if torch.cuda.is_available() else 'cpu')
# args.save = 'logs/search-{}'.format(args.save)
# if not os.path.exists(args.save):
# os.mkdir(args.save)
# log_filename = os.path.join(args.save, 'log.txt')
# init_logger('', log_filename, logging.INFO, False)
# print('*************log_filename=%s************' % log_filename)
# device = torch.device('cuda:1')
# device = torch.device('cpu')
print(f'here is the args{args}')
if not torch.cuda.is_available():
logging.info('no gpu device available')
sys.exit(1)
torch.cuda.set_device(args.gpu)
cudnn.benchmark = True
cudnn.enabled = True
set_seed(args.seed)
logging.info("args = %s", args.__dict__)
vocabulary_size = len(deserialize(f'{args.data_location}{args.catalog_name}_vocabulary.dat'))
embedding_width = vocabulary_size if args.embedding_size == 'one_hot' else int(args.embedding_size)
transform_to_emb_size = vocabulary_size if args.embedding_size == 'one_hot' else None
data_filename = f'{args.data_location}{args.catalog_name}_component_{args.embedding_size}_dgl.hdf5'
model_path = os.path.join(args.save_path, args.model_path)
model_output_file = os.path.join(model_path, f'{args.catalog_name}_{args.embedding_size}_{args.exp_id}_sub.dat')
model_plot_file = os.path.join(model_path, f'{args.catalog_name}_{args.embedding_size}_{args.exp_id}_sub.png')
args.in_dim = embedding_width + 7
args.out_dim = vocabulary_size + 7
split_dataset = [get_test_dataset(data_filename, f'data/prepare_{args.catalog_name}', transform_to_emb_size, mode=mode) for
mode in ['train', 'val', 'test']]
for i in range(len(['train', 'val', 'test'])):
split_dataset[i].graphs = [dgl.add_self_loop(graph) for graph in split_dataset[i].graphs]
fake_dataset = FakeDataset()
print(fake_dataset.tgraphs)
save_path = os.path.join(args.save_path, args.model_path, args.save_searched_file)
with open(save_path, 'rb') as f:
op_dict = pickle.load(f)
op_dict = {'FF': ['concat', 'concat', 'sum', 'mean'], 'NA': ['gatv2', 'gat', 'gin'], 'RE': ['global_max'], 'SC': ['identity', 'zero', 'identity', 'identity', 'identity', 'identity', 'identity', 'identity', 'zero', 'identity']}
# 'SC': ['identity', 'zero', 'identity', 'zero', 'zero', 'identity', 'identity', 'zero', 'zero', 'zero']}
# op_dict = {'LA': 'l_lstm', 'NA': ['gat', 'gat'], 'RE': ['global_att', 'global_att', 'global_att']}
# op_dict = {'LA': 'l_sum', 'NA': ['mlp', 'gin'], 'RE': ['set_transformer', 'set_transformer', 'set_transformer']}
# op_dict = {'LA': 'l_lstm', 'NA': ['gatv2'], 'RE': ['none', 'global_mean']}
print(f'search architecture {op_dict}')
model = SubNetwork(args, op_dict)
model = model.to(device)
logging.info("param size = %fMB", count_parameters_in_MB(model))
print(f"param size = {count_parameters_in_MB(model)}fMB")
trainer = SubFakePredictionTrainer(model, split_dataset[0], split_dataset[1], split_dataset[2], fake_dataset, model_output_file, args)
if not args.skip_training :
trainer.train(device)
evaluator = PredictionCsEvaluator(fake_dataset, trainer.model, model_output_file, args, device)
evaluator.evaluate(0, k_values_to_test=(1, 2, 3, 5, 10, 15, 20), investigate_attention=False)
logging.shutdown()
# def run_by_seed(args):
# res = []
# for i in range(args.num_sampled_archs):
# print('searched {}-th for {}...'.format(i+1, args.data))
# args.save = '{}-{}'.format(args.data, time.strftime("%Y%m%d-%H%M%S"))
# seed = np.random.randint(0, 10000)
# # seed = 0
# args.seed = seed
# genotype = main(args)
# res.append('seed={}, genotype={}, saved_dir={}'.format(seed, genotype, args.save))
# filename = './exp_res/%s-searched_res-%s-eps%s-reg%s.txt' % (args.data, time.strftime('%Y%m%d-%H%M%S'), args.epsilon, args.weight_decay)
# fw = open(filename, 'w+')
# fw.write('\n'.join(res))
# fw.close()
# print('searched res for {} saved in {}'.format(args.data, filename))
if __name__ == '__main__':
# if len(sys.argv) == 1:
# print('Please specify the required config info!!!')
# sys.exit(0)
args = get_args()
main(args)