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
# import logging
# logger = logging.getLogger()
# fh = logging.FileHandler('TrainLog/01_nodules.log',encoding='utf-8')
# fh.setLevel(logging.DEBUG)
# sh = logging.StreamHandler()
# sh.setLevel(logging.INFO)
# formatter = logging.Formatter('%(asctime)s %(filename)s[line:%(lineno)d] %(levelname)s %(message)s')
# fh.setFormatter(formatter)
# sh.setFormatter(formatter)
# logger.addHandler(fh)
# logger.addHandler(sh)
# logger.setLevel(10)
import os
import sys
import json
import numpy as np
from datetime import datetime
import torch
import torch.distributed
import torch.multiprocessing as mp
import torch.distributed as dist
import logging
import argparse
from Segmentation import Segmentation3D
sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
# sys.path.insert(0, os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(__file__))), ".."))
__package__ = "LungNoduleSegmentation"
from k8s_utils import CParamFiller, CK8sPathWrapper
# os.environ["CUDA_VISIBLE_DEVICES"] = "0,1,2,3,4,5,6,7"
import glob
def init_dist(backend='nccl',
master_ip='127.0.0.1',
port=29500):
if mp.get_start_method(allow_none=True) is None:
mp.set_start_method('spawn')
os.environ['MASTER_ADDR'] = master_ip
os.environ['MASTER_PORT'] = str(port)
rank = int(os.environ['RANK'])
world_size = int(os.environ['WORLD_SIZE'])
num_gpus = torch.cuda.device_count()
local_rank = os.environ['LOCAL_RANK']
deviceid = eval(local_rank) % num_gpus
torch.cuda.set_device(deviceid)
print(fr'dist settings: local_rank {local_rank}, rank {rank}, worldsize {world_size}, gpus {num_gpus}, deviceid {deviceid}')
dist.init_process_group(backend=backend)
return rank, world_size
def get_logger(name, task_name=None):
file_handler = logging.StreamHandler(sys.stdout)
logger = logging.getLogger(name)
logger.setLevel(logging.INFO)
logger.addHandler(file_handler)
return logger
def _log_msg(strmsg="\n"):
if torch.distributed.get_rank() == 0:
if g_logger is not None:
g_logger.info(strmsg)
else:
print(strmsg)
if __name__ == '__main__':
parser = argparse.ArgumentParser(description="full functional execute script of Segmentation module.")
group = parser.add_mutually_exclusive_group()
# ddp
parser.add_argument("--local_rank", default=-1, type=int)
parser.add_argument('--port', default=29500, type=int, help='port of server')
parser.add_argument('--world_size', default=1, type=int)
parser.add_argument('--rank', default=0, type=int)
parser.add_argument('--master_ip', default='127.0.0.1', type=str)
parser.add_argument('--job_data_root', default="/root/Documents/GroupLung/Datasets/seg_data/job_data_seg_train", type=str)
args = parser.parse_args()
g_pathWrapper = CK8sPathWrapper(args.job_data_root)
config_path = g_pathWrapper.get_input_inputconfig_filepath()
cfg = CParamFiller.fillin_args(config_path, args) # config_path
train_params = 'configs/train/input_config.json'
cfg = CParamFiller.fillin_args(train_params, cfg)
g_logger = get_logger(__name__)
rank, world_size = init_dist(backend='nccl', master_ip=args.master_ip, port=args.port)
cfg.rank = rank
cfg.world_size = world_size
cfg.local_rank = os.environ['LOCAL_RANK']
if cfg.pretrain_msg:
cfg.pretrain_msg = os.path.join(g_pathWrapper.get_input_dirpath(), cfg.pretrain_msg)
print(cfg)
cfg.common['train_data_path'] = os.path.join(g_pathWrapper.get_output_tmp_dirpath(), 'train_split.json')
# cfg.common['train_mask_path'] = os.path.join(g_pathWrapper.get_output_tmp_dirpath(), 'train_mask.npy')
# cfg.common['train_info_path'] = os.path.join(g_pathWrapper.get_output_tmp_dirpath(), 'train_info.npy')
cfg.common['val_data_path'] = os.path.join(g_pathWrapper.get_output_tmp_dirpath(), 'val_split.json')
# cfg.common['val_mask_path'] = os.path.join(g_pathWrapper.get_output_tmp_dirpath(), 'val_mask.npy')
# cfg.common['val_info_path'] = os.path.join(g_pathWrapper.get_output_tmp_dirpath(), 'val_info.npy')
train_rate = cfg.split['train']
if (not os.path.exists(cfg.common['train_data_path'])) and cfg.rank == 0:
# 读取所有的npy信息，并划分训练集和测试集
# whole_data = []
# whole_mask = []
# whole_info = []
npy_path = g_pathWrapper.get_output_preprocess_npy_path()
data_list = glob.glob(os.path.join(npy_path, '*', '*_data.npy'))
all_len = len(data_list)
train_len = int(all_len * train_rate)
train_list = data_list[:train_len]
val_list = data_list[train_len:]
train_data = {'data': train_list}
val_data = {'data': val_list}
with open(cfg.common['train_data_path'], 'w+') as f:
json.dump(train_data, f, indent=4)
with open(cfg.common['val_data_path'], 'w+') as f:
json.dump(val_data, f, indent=4)
# for img_path in glob.glob(os.path.join(npy_path, '*', '*_data.npy')):
# mask_path = img_path.replace('_data.npy', '_mask.npy')
# info_path = img_path.replace('_data.npy', '_info.npy')
# whole_data.append(np.load(img_path)[None, ...])
# whole_mask.append(np.load(mask_path)[None, ...])
# whole_info.append(np.load(info_path)[None, ...])
# whole_data = np.concatenate(whole_data, axis=0)
# whole_mask = np.concatenate(whole_mask, axis=0)
# whole_info = np.concatenate(whole_info, axis=0)
# print(whole_data.shape, whole_mask.shape, whole_info.shape)
# # get train and test npy and info, mask
# train_data = []
# val_data = []
# train_mask = []
# val_mask = []
# train_info = []
# val_info = []
# indices = range(whole_data.shape[0])
# train_idxs = list(np.random.choice(indices, int(whole_data.shape[0] * train_rate), replace=False))
# val_idxs = [idx for idx in indices if idx not in train_idxs]
# for idx in train_idxs:
# train_data.append(whole_data[idx][np.newaxis,:])
# train_mask.append(whole_mask[idx][np.newaxis,:])
# train_info.append(whole_info[idx][np.newaxis,:])
# for idx in val_idxs:
# val_data.append(whole_data[idx][np.newaxis,:])
# val_mask.append(whole_mask[idx][np.newaxis,:])
# val_info.append(whole_info[idx][np.newaxis,:])
# train_data = np.concatenate(train_data, axis=0)
# train_info = np.concatenate(train_info, axis=0)
# train_mask = np.concatenate(train_mask, axis=0)
# val_data = np.concatenate(val_data, axis=0)
# val_info = np.concatenate(val_info, axis=0)
# val_mask = np.concatenate(val_mask, axis=0)
# np.save(cfg.common['train_data_path'], train_data)
# np.save(cfg.common['train_mask_path'], train_mask)
# np.save(cfg.common['train_info_path'], train_info)
# np.save(cfg.common['val_data_path'], val_data)
# np.save(cfg.common['val_mask_path'], val_mask)
# np.save(cfg.common['val_info_path'], val_info)
# torch.distributed.barrier()
# 设置输出的文件路径
cfg.common['base_path'] = g_pathWrapper.get_output_train_dirpath()
cfg.common['save_path'] = g_pathWrapper.get_output_train_bestmodel_dirpath()
cfg.common['writer_path'] = g_pathWrapper.get_output_train_writer_path()
cfg.common['eval_pmd'] = g_pathWrapper.get_output_eval_performance_md_filepath()
cfg.common['eval_pjson'] = g_pathWrapper.get_output_eval_performance_filepath()
cfg.common['train_metrics'] = g_pathWrapper.get_output_train_trainingmetrics_filepath()
cfg.common['train_result'] = g_pathWrapper.get_output_train_trainresult_filepath()
if cfg.rank == 0:
train_result_dicts = {'successFlag': '', 'bestModelEpoch': 0}
with open(cfg.common['train_result'], 'w+') as file:
json.dump(train_result_dicts, file, indent=4)
# 开始训练, mode choice in [training, testing]
train_obj = Segmentation3D(cfg, g_logger)
train_obj()