-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathtrain_mul.py
executable file
·134 lines (104 loc) · 5.61 KB
/
train_mul.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
import sys
import os
from torch.utils.data import DataLoader
import configargparse
import time
import numpy as np
import shutil
from IPython import embed
import training, modules, loss_functions
import torch
import utils
from torch.utils.tensorboard import SummaryWriter
from tqdm.autonotebook import tqdm
import dataset, modify_points, utils_eval
import dataio
# import wandb
p = configargparse.ArgumentParser()
# General training options
p.add_argument('--dataset_path', type=str, default='/home/user/pyProject/dataset/odometry_velodyne/05/newfile/')
p.add_argument('--log_path', type=str, default='logs/')
p.add_argument('--expr_name', type=str, default='test')
p.add_argument('--integtxt', type=str, default='/home/user/pyProject/siren_mlp_en/data/integtxt.txt')
p.add_argument('--logging_root', type=str, default='/home/user/pyProject/siren_mlp_en/logs/', help='root for logging')
p.add_argument('--experiment_name', type=str, default='experiment05_fourier')
# p.add_argument('--experiment_name', type=str, default='kitti_label_experiment_fourier')
p.add_argument('--backup', type=str, default='kitti05_backup_fourier')
# p.add_argument('--backup', type=str, default='kitti_label_backup_fourier')
p.add_argument('--num_epochs', type=int, default=100,
help='Number of epochs to train for.')
p.add_argument('--epochs_til_ckpt', type=int, default=100,
help='Time interval in seconds until checkpoint is saved.')
p.add_argument('--steps_til_summary', type=int, default=100,
help='Time interval in seconds until tensorboard summary is saved.')
p.add_argument('--clip_grad', type=bool, default=True)
p.add_argument('--model_type', type=str, default='sine',
help='Options are "sine" (all sine activations) and "mixed" (first layer sine, other layers tanh)')
p.add_argument('--eval_resolution', type=int, default=256)
p.add_argument('--data_class', type=str, default='all',
help='Options are "all"(train & valid & test) and "valid"')
####################################################################
p.add_argument('--batch_size', type=int, default=160000) # the minimum count of points is 4995 (10, 000195)
p.add_argument('--lr', type=float, default=1e-4, help='learning rate.')
p.add_argument('--fix_coordiante', type=bool, default=True)
p.add_argument('--pemode', type=str, default='fourier') ## nerf mlp fourier
p.add_argument('--num_encoding_functions', type=int, default=10)
####################################################################
p.add_argument('--incremental_sampling', type=bool, default=True)
p.add_argument('--pre_sample_rate', type=float, default=0.75)
####################################################################
p.add_argument('--frames', type=int, default=920)
p.add_argument('--interval', type=int, default=10)
p.add_argument('--frame_accumulate_num', type=int, default=10)
#################################################################
p.add_argument('--use_pre_model', type=bool, default=False)
p.add_argument('--pthpath', type=str, default='/home/user/pyProject/siren_mlp_en/logs/kitti05_backup_fourier/440.pth')
opt = p.parse_args()
# wandb.init(project="siren")
model = modules.semanticSIREN(type=opt.model_type,in_features=3,pemode=opt.pemode,num_encoding_functions=opt.num_encoding_functions)
if opt.use_pre_model:
model.load_state_dict(torch.load(opt.pthpath))
model.cuda()
loss_fn = loss_functions.sdf_label
summary_fn = utils.write_sdf_summary
root_path = os.path.join(opt.logging_root, opt.experiment_name)
if not os.path.exists(root_path):
os.makedirs(root_path)
if not opt.use_pre_model:
if os.path.exists(opt.integtxt):
os.remove(opt.integtxt)
if not opt.use_pre_model:
first_frame=True
else:
first_frame=False
pre_size=0
for index in range(450,opt.frames,opt.interval):
# time1 = time.time()
input_data_path=open(opt.integtxt,'a+')
if ((opt.incremental_sampling==True) and (first_frame == False)):
pc = np.genfromtxt(opt.integtxt)
pre_size=pc.shape[0]
print("pre_size",pre_size)
num=index
for i in range(0,opt.frame_accumulate_num,1):
filepath=os.path.join(opt.dataset_path,str((num+i)).zfill(6)+".txt") ## for kitti dataset,1 frame of every 5 frames
# filepath=os.path.join(opt.dataset_path,str((num+i)*10)+".xyz") ##for ICL dataset
f=open(filepath)
input_data_path.write(f.read()+'\n')
print(filepath)
input_data_path.close()
train_dataset = dataio.PointCloud(opt.integtxt,on_surface_points=opt.batch_size,fix_coordiante=opt.fix_coordiante,incremental_sampling=opt.incremental_sampling,pre_size=pre_size,first_frame=first_frame,pre_sample_rate=opt.pre_sample_rate)
train_dataloader = DataLoader(train_dataset, shuffle=True, batch_size=1, pin_memory=True, num_workers=16)
changingepoch=10
if not os.path.exists(os.path.join(opt.logging_root,opt.backup)):
os.makedirs(os.path.join(opt.logging_root,opt.backup))
backuppth= os.path.join(opt.logging_root,opt.backup)
training.train(index,model=model, train_dataloader=train_dataloader, epochs=changingepoch, lr=opt.lr,
steps_til_summary=opt.steps_til_summary, epochs_til_checkpoint=opt.epochs_til_ckpt,
model_dir=root_path, backuppth=backuppth, loss_fn=loss_fn, summary_fn=summary_fn, double_precision=False,
clip_grad=True)
first_frame=False
# time2 = time.time()
# print('time elapse:', (time2-time1)//60, 'min', (time2-time1)%60, 's')
torch.cuda.empty_cache()
os.remove(opt.integtxt) ##文件一直追加数据，所以最后删除