def local_tail(self, logfile, webuser):
# 创建一个可跨文件的全局变量，以便控制死循环
gl._init()
gl.set_value('deploy_' + str(webuser), False)
try:
with open(logfile, 'rt') as f:
f.seek(0, 0)
while True:
is_stop = gl.get_value('deploy_' + str(webuser))
line = f.readline()
if line:
self.send_message(webuser, line)
elif is_stop:
self.send_message(webuser, '[INFO]文件监视结束..')
break
except Exception as e:
self.send_message(webuser, e)
def remote_tail(self,
host,
port,
user,
passwd,
logfile,
webuser,
filter_text=None):
# 创建一个可跨文件的全局变量，控制停止
try:
self.client = paramiko.SSHClient()
self.client.load_system_host_keys()
self.client.set_missing_host_key_policy(paramiko.AutoAddPolicy())
self.client.connect(hostname=host,
port=port,
username=user,
password=passwd)
interact = SSHClientInteraction(self.client,
timeout=10,
display=False)
interact.expect('.*#.*')
logfile = logfile.strip().replace('&&',
'').replace('||',
'').replace('|', '')
self.send_message(webuser, '[INFO][%s@%s]开始监控日志' % (user, host))
gl._init()
gl.set_value('tail_' + str(webuser), self.client)
if filter_text:
filter_text_re = filter_text.strip().replace('&&', '').replace(
'||', '').replace('|', '')
interact.send('tail -f %s|grep --color=never %s' %
(logfile, filter_text_re))
else:
interact.send('tail -f %s' % (logfile))
interact.tail(
output_callback=lambda m: self.send_message(webuser, m))
except Exception as e:
self.send_message(webuser, e)
finally:
try:
self.client.close()
except Exception as e:
self.send_message(webuser, e)
from model import Discriminator, resnet_56, resnet_56_sparse
from data import cifar10
from resnet import ResNet18, ResNet50
from resnet_sprase import ResNet18_sprase, ResNet50_sprase
from collections import OrderedDict
import numpy as np
from torch.autograd import Variable
from resnet_imagenet import resnet101
import torchvision.datasets as datasets
import torchvision.transforms as transforms
import utils.globalvar as gl
gl._init()
import time
import logging
import sys
try:
import nvidia.dali.plugin.pytorch as plugin_pytorch
from nvidia.dali.pipeline import Pipeline
import nvidia.dali.ops as ops
import nvidia.dali.types as types
except ImportError:
raise ImportError(
"Please install DALI from https://www.github.com/NVIDIA/DALI to run this example."
)
num_gpu = 4
help='Use multi-processing distributed training to launch '
'N processes per node, which has N GPUs. This is the '
'fastest way to use PyTorch for either single node or '
'multi node data parallel training')
parser.add_argument('--channel_removed_ratio',default=0.2,type=float,help='removed ratio.')
parser.add_argument('--spatial_removed_ratio',default=0.2,type=float,help='removed ratio.')
parser.add_argument('--Is_spatial',action='store_true',help='use spatial module or not,default is channel with conv.')
parser.add_argument('--lasso',action='store_true',help='add l1 regularization to channel module.')
parser.add_argument('--l1_coe',default=1e-8,type=float,help='coe of l1 regularization.')
parser.add_argument('--show',action='store_true',help='show model architecture.')
parser.add_argument('--flops',action='store_true',help='calc flops given a pretrained model.')
parser.add_argument('--debug',action='store_true',help='debug.')
args = parser.parse_args()
os.environ["CUDA_VISIBLE_DEVICES"] = str(args.gpu)
best_acc1 = 0
gvar._init()
gvar.set_value('removed_ratio_c',args.channel_removed_ratio)
gvar.set_value('removed_ratio_s',args.spatial_removed_ratio)
gvar.set_value('is_spatial',args.Is_spatial)
def main():
if not os.path.isdir(args.save_dir):
os.makedirs(args.save_dir)
if args.seed is not None:
random.seed(args.seed)
torch.manual_seed(args.seed)
cudnn.deterministic = True
warnings.warn('You have chosen to seed training. '
'This will turn on the CUDNN deterministic setting, '
'which can slow down your training considerably! '
'You may see unexpected behavior when restarting '
def post(self, request, format=None):
if request.data['excu'] == 'init':
# 项目初始化
id = request.data['id']
result = self.repo_init(id)
if result.exited == 0:
Project.objects.filter(id=id).update(status='Succeed')
info_logger.info('初始化项目:' + str(id) + ',执行成功!')
http_status = OK
msg = '初始化成功!'
else:
error_logger.error('初始化项目:%s 执行失败! 错误信息:%s' %
(str(id), result.stderr))
http_status = BAD
msg = '初始化项目:%s 执行失败! 错误信息:%s' % (str(id), result.stderr)
return XopsResponse(msg, status=http_status)
elif request.data['excu'] == 'deploy':
# 部署操作
id = request.data['id']
webuser = request.user.username
alias = request.data['alias']
self.start_time = time.strftime("%Y%m%d%H%M%S", time.localtime())
record_id = str(alias) + '_' + str(self.start_time)
name = '部署_' + record_id
DeployRecord.objects.create(name=name,
alias=alias,
status='Failed',
project_id=int(id))
Project.objects.filter(id=id).update(last_task_status='Failed')
local_log_path = self._path.rstrip('/') + '/' + str(
id) + '_' + str(request.data['alias']) + '/logs'
log = local_log_path + '/' + record_id + '.log'
version = request.data['version'].strip()
serverid = request.data['server_ids']
deploy = DeployExcu(webuser, record_id, id)
deploy.start(log, version, serverid, record_id, webuser,
self.start_time)
return XopsResponse(record_id)
elif request.data['excu'] == 'rollback':
# 回滚
id = request.data['id']
project_id = request.data['project_id']
alias = request.data['alias']
self.start_time = time.strftime("%Y%m%d%H%M%S", time.localtime())
record_id = str(alias) + '_' + str(self.start_time)
log = self._path.rstrip('/') + '/' + str(project_id) + '_' + str(
alias) + '/logs/' + record_id + '.log'
self.do_rollback(id, log, record_id)
return XopsResponse(record_id)
elif request.data['excu'] == 'deploymsg':
# 部署控制台消息读取
try:
id = request.data['id']
alias = request.data['alias']
record = request.data['record']
scenario = int(request.data['scenario'])
logfile = self._path.rstrip('/') + '/' + str(id) + '_' + str(
alias) + '/logs/' + record + '.log'
webuser = request.user.username
print(webuser)
msg = Tailf()
if scenario == 0:
gl._init()
gl.set_value('deploy_' + str(webuser), False)
msg.local_tailf(logfile, webuser)
http_status = OK
request_status = '执行成功!'
except Exception as e:
http_status = BAD
request_status = '执行错误:日志文件可能不存在!'
print(e)
return XopsResponse(request_status, status=http_status)
elif request.data['excu'] == 'readlog' and request.data[
'scenario'] == 1:
# 读取部署日志
try:
id = request.data['id']
alias = request.data['alias']
record = request.data['record']
logfile = self._path.rstrip('/') + '/' + str(id) + '_' + str(
alias) + '/logs/' + record + '.log'
response = FileResponse(open(logfile, 'rb'))
response['Content-Type'] = 'text/plain'
return response
except Exception:
http_status = BAD
request_status = '执行错误:文件不存在!'
return XopsResponse(request_status, status=http_status)
elif request.data['excu'] == 'app_start':
# 项目启动
try:
app_start = request.data['app_start']
host = request.data['host']
webuser = request.user.username
auth_info, auth_key = auth_init(host)
connect = Shell(auth_info,
connect_timeout=5,
connect_kwargs=auth_key)
app_start = app_start.strip().replace('&&',
'').replace('||', '')
connect.run(app_start, ws=True, webuser=webuser)
connect.close()
http_status = OK
request_status = '执行成功!'
except Exception as e:
http_status = BAD
request_status = '执行错误:' + str(e)
return XopsResponse(request_status, status=http_status)
elif request.data['excu'] == 'app_stop':
# 项目停止
try:
app_stop = request.data['app_stop']
host = request.data['host']
webuser = request.user.username
auth_info, auth_key = auth_init(host)
connect = Shell(auth_info,
connect_timeout=5,
connect_kwargs=auth_key)
app_stop = app_stop.strip().replace('&&', '').replace('||', '')
connect.run(app_stop, ws=True, webuser=webuser)
connect.close()
http_status = OK
request_status = '执行成功!'
except Exception as e:
http_status = BAD
request_status = '执行错误:' + str(e)
return XopsResponse(request_status, status=http_status)
elif request.data['excu'] == 'tail_start':
# 日志监控
try:
filter_text = str(request.data['filter'])
app_log_file = request.data['app_log_file']
host = request.data['host']
webuser = request.user.username
device_info = DeviceInfo.objects.filter(id=int(host)).values()
host = device_info[0]['hostname']
auth_type = device_info[0]['auth_type']
connect_info = ConnectionInfo.objects.filter(
hostname=host, auth_type=auth_type).values()
user = connect_info[0]['username']
passwd = connect_info[0]['password']
port = connect_info[0]['port']
tail = Tailf()
tail.remote_tail(host,
port,
user,
passwd,
app_log_file,
webuser,
filter_text=filter_text)
http_status = OK
request_status = '执行成功!'
except Exception as e:
http_status = BAD
request_status = str(e)
return XopsResponse(request_status, status=http_status)
elif request.data['excu'] == 'tail_stop':
# 日志监控停止
try:
webuser = request.user.username
if hasattr(gl, '_global_dict'):
tail_key = 'tail_' + str(webuser)
if tail_key in gl._global_dict.keys():
client = gl.get_value('tail_' + str(webuser))
client.close()
http_status = OK
request_status = '执行成功!'
except Exception as e:
http_status = BAD
request_status = str(e)
return XopsResponse(request_status, status=http_status)