/
train_origami.py
executable file
·120 lines (95 loc) · 3.95 KB
/
train_origami.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
import argparse
import datetime
import os
import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
from tqdm import tqdm
from hodgenet import HodgeNetModel
from meshdata import OrigamiDataset
def main(args):
torch.set_default_dtype(torch.float64) # needed for eigenvalue problems
torch.manual_seed(0)
np.random.seed(0)
dataset = OrigamiDataset(
edge_features_from_vertex_features=['vertices'],
triangle_features_from_vertex_features=['vertices'])
def mycollate(b): return b
dataloader = DataLoader(dataset, batch_size=args.bs,
num_workers=0, collate_fn=mycollate)
example = dataset[0]
hodgenet_model = HodgeNetModel(
example['int_edge_features'].shape[1],
example['triangle_features'].shape[1],
num_output_features=args.n_out_features, mesh_feature=True,
num_eigenvectors=args.n_eig, num_extra_eigenvectors=args.n_extra_eig,
resample_to_triangles=False,
num_bdry_edge_features=example['bdry_edge_features'].shape[1],
num_vector_dimensions=args.num_vector_dimensions)
origami_model = nn.Sequential(
hodgenet_model,
nn.Linear(args.n_out_features*args.num_vector_dimensions *
args.num_vector_dimensions, 32),
nn.LayerNorm(32),
nn.LeakyReLU(),
nn.Linear(32, 16),
nn.LayerNorm(16),
nn.LeakyReLU(),
nn.Linear(16, 2))
optimizer = optim.AdamW(origami_model.parameters(), lr=args.lr)
if not os.path.exists(args.out):
os.makedirs(args.out)
train_writer = SummaryWriter(os.path.join(
args.out, datetime.datetime.now().strftime('train-%m%d%y-%H%M%S')),
flush_secs=1)
def epoch_loop(dataloader, epochname, epochnum, writer, optimize=True):
epoch_loss, epoch_size = 0, 0
pbar = tqdm(total=len(dataloader),
desc='{} {}'.format(epochname, epochnum))
for batchnum, batch in enumerate(dataloader):
if optimize:
optimizer.zero_grad()
batch_loss = 0
dirs = origami_model(batch)
dirs = F.normalize(dirs, p=2, dim=-1)
for mesh, dir_estimate in zip(batch, dirs):
gt_dir = mesh['dir'].to(dir_estimate.device)
batch_loss += 1 - (gt_dir * dir_estimate).sum(-1)
batch_loss /= len(batch)
pbar.set_postfix({
'loss': batch_loss.item(),
})
pbar.update(1)
epoch_loss += batch_loss.item()
epoch_size += 1
if optimize:
batch_loss.backward()
nn.utils.clip_grad_norm_(origami_model.parameters(), 1)
optimizer.step()
writer.add_scalar('Loss', batch_loss.item(),
epochnum*len(dataloader)+batchnum)
pbar.close()
for epoch in range(args.n_epochs):
origami_model.train()
epoch_loop(dataloader, 'Epoch', epoch, train_writer)
torch.save({
'origami_model_state_dict': origami_model.state_dict(),
'opt_state_dict': optimizer.state_dict(),
'epoch': epoch
}, os.path.join(args.out, f'{epoch}.pth'))
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--out', type=str, default='out/origami')
parser.add_argument('--bs', type=int, default=16)
parser.add_argument('--lr', type=float, default=1e-3)
parser.add_argument('--n_epochs', type=int, default=10000)
parser.add_argument('--n_eig', type=int, default=32)
parser.add_argument('--n_extra_eig', type=int, default=32)
parser.add_argument('--n_out_features', type=int, default=32)
parser.add_argument('--num_vector_dimensions', type=int, default=4)
args = parser.parse_args()
main(args)