-
Notifications
You must be signed in to change notification settings - Fork 1
/
train.py
170 lines (131 loc) · 6.34 KB
/
train.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
from encodings import normalize_encoding
from locale import normalize
import os
import sys
import time
import random
import pandas as pd
import shutil
import networkx as nx
import argparse
import json
from typing import List
import numpy as np
import tensorflow as tf
os.environ['TF_CPP_MIN_LOG_LEVEL'] = "3"
from methods.HNN_HM.hnn_hm.hypergraph import HypergraphsTuple
from methods.HNN_HM.hnn_hm.models_trust import HypergraphModelTrust3
from methods.HNN_HM.hnn_hm.config.artery import ArteryConfig
from utils import artery, artery_eval
from tqdm import tqdm
np.set_printoptions(suppress=True)
def set_gpu_memory_growth():
gpus = tf.config.experimental.list_physical_devices('GPU')
for gpu in gpus:
tf.config.experimental.set_memory_growth(gpu, True)
def set_random_seed(seed):
random.seed(seed)
np.random.seed(seed)
tf.random.set_seed(seed)
def get_optimizer(optimizer_name, lr, momentum):
assert optimizer_name in ['adam', 'sgd']
if optimizer_name == 'adam':
optimizer = tf.keras.optimizers.Adam(lr, amsgrad=True)
elif optimizer_name == 'sgd':
optimizer = tf.keras.optimizers.SGD(learning_rate=lr, momentum=momentum)
return optimizer
class TrueClassProbabilityLoss(tf.keras.losses.Loss):
def __init__(self):
super(TrueClassProbabilityLoss, self).__init__()
def call(self, y_true, tcp_conf_pred):
tcp_conf = tcp_conf_pred[:, :1]
tcp_pred = tcp_conf_pred[:, 1:]
p_target = tf.gather(tcp_pred, y_true, axis=1, batch_dims=1)
p_target = tf.reshape(p_target, [-1])
confidence_loss = tf.losses.mean_squared_error(tcp_conf, p_target)
return tf.reduce_mean(confidence_loss)
def tcp_loss_fn(y_true, tcp_conf, tcp_pred):
p_target = tf.gather(tcp_pred, y_true, axis=1, batch_dims=1)
p_target = tf.reshape(p_target, [-1])
confidence_loss = tf.losses.mean_squared_error(tcp_conf, p_target)
return tf.reduce_mean(confidence_loss)
class HGM_Trainer(object):
def __init__(self, params) -> None:
self.params = params
self.rand = np.random.RandomState(seed=params.seed)
set_random_seed(args.seed)
os.environ["CUDA_VISIBLE_DEVICES"] = str(self.params.gpu)
set_gpu_memory_growth()
self.__init_database__()
self.__init_dataloader__()
self.__init_model__()
def __init_database__(self):
dataset, _ = artery._load_graph_in_mem(args.data_file_path)
self.dataset = dataset
def __init_model__(self):
model = HypergraphModelTrust3(num_processing_steps=self.params.num_message_passing,
node_output_size=16,
num_hidden_layer=self.params.num_layers,
hidden_layer_size=self.params.latent_size,
node_attention=self.params.node_attention,
edge_attention=self.params.edge_attention)
self.model = model
def __init_dataloader__(self):
problem_generator_test = artery.ArteryProblemGeneratorTest(self.rand, self.dataset, self.dataset, self.params.cache, self.params.cache_path)
cfg = ArteryConfig(self.params.feature_dim)
graph_creator = artery.ArteryGraphCreator(cfg)
dataloader_test = artery.DataLoader(problem_generator_test, graph_creator)
self.dataloader_test = dataloader_test
def __restore__(self):
print(f"HGM_Trainer.__restore__: {self.params.exp}")
self.model.load_weights(f"{self.params.exp}/weights.ckpt")
def demo(self):
save_path = "demo_output"
test_sample_id, template_sample_id = "test1_RAO_CAU", "template1_RAO_CAU"
g0, g1 = self.dataset[test_sample_id]['g'], self.dataset[template_sample_id]['g']
n0, n1 = g0.number_of_nodes(), g1.number_of_nodes()
inputs_te, targets_te, ahg = self.dataloader_test.generate_data_test(test_sample_id, template_sample_id)
outputs, output_confs, _ = self.model(inputs_te)
outputs = outputs[-1].nodes.numpy().reshape(n0, n1)
output_confs = output_confs[-1].nodes.numpy().reshape(n0, n1)
gt = targets_te.nodes.numpy().reshape(n0, n1)
mappings, total, matched, unmatched, output_hug = artery_eval.compute_accuracy_artery_node2(gt, outputs, self.dataset, ahg, save_path, self.params.plot)
print(mappings)
if __name__ == '__main__':
parser = argparse.ArgumentParser()
# data
parser.add_argument('--data_file_path', type=str, default="data/artery_with_feature")
parser.add_argument('--cache_path', type=str, default="cache")
parser.add_argument('--cache', type=bool, default=True)
parser.add_argument('--gpu', type=int, default=0)
parser.add_argument('--exp', type=str, default="saved_models")
parser.add_argument('--cv', type=int, default=0)
parser.add_argument('--cv_max', type=int, default=5)
parser.add_argument('--template_ratio', type=float, default=0.1)
parser.add_argument('--seed', type=int, default=1234)
parser.add_argument('--feature_dim', type=int, default=121)
parser.add_argument('--train', type=str, default="demo")
parser.add_argument('--alpha', type=float, default=0.1)
parser.add_argument('--trust_threshold', type=float, default=0.5)
# model
parser.add_argument('--batch_size_tr', type=int, default=4)
parser.add_argument('--node_attention', type=bool, default=True)
parser.add_argument('--edge_attention', type=bool, default=True)
parser.add_argument('--num_workers', type=int, default=1)
parser.add_argument('--batch_size_te', type=int, default=1)
parser.add_argument('--num_message_passing', type=int, default=2)
parser.add_argument('--num_layers', type=int, default=2)
parser.add_argument('--latent_size', type=int, default=64)
# training
parser.add_argument('--n_train_iterations', type=int, default=100001)
parser.add_argument('--eval_interval', type=int, default=2000)
parser.add_argument('--plot', type=bool, default=True)
# optimizer
parser.add_argument('--learning_rate', type=float, default=0.0001)
parser.add_argument('--optimizer', type=str, default="adam")
parser.add_argument('--momentum', type=float, default=0.9)
args = parser.parse_args()
assert args.train == "demo"
hgm_trainer = HGM_Trainer(args)
hgm_trainer.__restore__()
hgm_trainer.demo()