-
Notifications
You must be signed in to change notification settings - Fork 8
/
helpers.py
203 lines (151 loc) · 6.7 KB
/
helpers.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
import torch
from torch.utils.data import Dataset
import numpy as np
from pyFiles import LDraw, LegoGraph, utils
from DGL_GIN.gin import GIN
import os
import ast
import random
import time
import pickle
import json
class FileToTarget():
classes = {}
def get_target(self, filename):
class_name = filename.split('_')[0]
if class_name not in self.classes:
print(class_name)
self.classes[class_name] = len(self.classes)
return self.classes[class_name]
def get_class_name(self, class_num):
return list(self.classes.keys())[class_num]
def num_classes(self):
return len(self.classes)
def get_all_classes(self):
return list({k: v for k, v in sorted(self.classes.items(), key=lambda item: item[1])}.keys())
def include_in_dataset(filename, config = None):
is_ldr = filename.endswith('.ldr')
splits = filename.split('_')
class_name = splits[0]
augmented = splits[1]
if is_ldr and (class_name != '2blocks-perpendicular' \
and class_name != '2blocks' and class_name != 'random'):
if 'augmented' not in augmented or config['include_augmented']:
return True
return False
class LegoDataset(Dataset):
def __init__(self, config = None):
super().__init__()
self.dataset = []
self.file_to_target = FileToTarget()
try:
dataset_with_filenames = self.load_dataset_from_file(config)
print('loaded dataset from file')
except FileNotFoundError:
print('making dataset')
dataset_with_filenames = self.make_dataset_from_LDraw(config)
self.write_dataset_to_file(config, dataset_with_filenames)
self.add_dataset(dataset_with_filenames)
random.shuffle(self.dataset)
def load_dataset_from_file(self, config):
pickle_file = 'dataset/augmented-{}.dat'.format(config['include_augmented'])
with open(pickle_file, 'rb') as f:
dataset_with_filenames = pickle.load(f)
return dataset_with_filenames
def add_dataset(self, dataset_with_filenames):
to_sequence = utils.LegoGraphToActionSequence()
for graph_with_filename in dataset_with_filenames:
filename = graph_with_filename[0]
graph = graph_with_filename[1]
self.add_graph_to_dataset(graph, filename, to_sequence)
def make_dataset_from_LDraw(self, config):
dataset_with_filenames = []
to_sequence = utils.LegoGraphToActionSequence()
directory = os.fsencode(os.path.join(os.getcwd() + '/ldr_files/dataset'))
for file in sorted(os.listdir(directory)):
filename = os.fsdecode(file)
directory_str = os.fsdecode(directory) + '/'
if include_in_dataset(filename, config = config):
lg = LDraw.LDraw_to_graph(directory_str + filename)
dataset_with_filenames.append([filename, lg])
return dataset_with_filenames
def write_dataset_to_file(self, config, dataset_with_filenames):
try:
os.mkdir(os.path.join(os.getcwd(), 'dataset'))
except FileExistsError:
pass
pickle_file = 'dataset/augmented-{}.dat'.format(config['include_augmented'])
with open(pickle_file, 'wb') as f:
pickle.dump(dataset_with_filenames, f)
def __len__(self):
return len(self.dataset)
def collate_single(self, batch):
assert len(batch) == 1, 'Currently we do not support batched training'
return batch[0]
def collate_batch(self, batch):
return batch
class GetDGMGDataset(LegoDataset):
def __init__(self, config = None):
super().__init__(config = config)
def add_graph_to_dataset(self, graph, filename, graph_to_sequence):
class_num = self.file_to_target.get_target(filename)
actions = graph_to_sequence.to_action_sequence(graph, class_num)
actions.insert(0, filename)
self.dataset.append(actions)
def __getitem__(self, index):
return self.dataset[index]
class GINDataset(LegoDataset):
def __init__(self, list_of_graphs = None, with_edge_types = False, config = None):
self.with_edge_types = with_edge_types
if list_of_graphs is None:
super().__init__(config = config)
elif list_of_graphs is not None:
self.dataset = list_of_graphs
self.gclasses = len(FileToTarget().classes)
def add_graph_to_dataset(self, graph, filename, *args):
try:
del graph.ndata['hg']
except:
pass
if 'attr' not in graph.ndata:
add_node_attributes(graph)
if self.with_edge_types:
add_edge_attributes(graph)
class_num = self.file_to_target.get_target(filename)
if class_num >= 8:
class_num += 1
self.dataset.append({'graph': graph, 'target': class_num, 'filename': filename})
def __getitem__(self, index):
return self.dataset[index]['graph'], self.dataset[index]['target']
def add_edge_attributes(lg):
eye = torch.eye(49)
embed_edge = lambda x_shift, z_shift: eye[x_shift + 3 + ((z_shift + 3) * 7)]
init_attr_graph = torch.Tensor([])
for dest in range(lg.number_of_nodes()):
init_attr_node = torch.zeros(49)
for src in lg.in_edges(dest)[0].detach().numpy():
shift = ast.literal_eval(lg.edge_labels[src, dest])
init_attr_node += embed_edge(shift[0], shift[1])
init_attr_node = torch.cat([lg.ndata['attr'][dest], init_attr_node]).view(-1, 51)
init_attr_graph = torch.cat([init_attr_graph, init_attr_node])
lg.ndata['attr'] = init_attr_graph
def add_node_attributes(lg):
eye = np.eye(2)
embed_node = lambda x: eye[1] if x == 'Brick(4, 2)' else eye[0]
init_attrs = torch.tensor([embed_node(lg.node_labels[x]) for x in range(lg.number_of_nodes())], dtype = torch.float32)
lg.ndata['attr'] = init_attrs
def save_model(model, epoch, dir, optimizer, scheduler = None):
filename = os.path.join(dir, 'models', 'epoch_{:04d}.h5'.format(epoch))
to_save = {'model_state_dict': model.state_dict(),
'optimizer_state_dict': optimizer.state_dict()}
if scheduler is not None:
to_save['scheduler_state_dict'] = scheduler.state_dict()
torch.save(to_save, filename)
def load_gin():
file = torch.load('pretrained_GIN.h5')
config = file['gin_config']
gin = GIN(config['num_layers'], config['num_mlp_layers'], config['input_dim'],
config['hidden_dim'], config['output_dim'], config['final_dropout'],
config['learn_eps'], config['graph_pooling_type'], config['neighbor_pooling_type'])
gin.load_state_dict(file['gin_state_dict'])
return gin