# total number of parameters
param_count = 0
for param in model.parameters():
param_count += param.view(-1).size()[0]
if not os.path.exists(config.log):
os.mkdir(config.log)
if opt.log == '':
log_path = config.log + utils.format_time(time.localtime()) + '/'
else:
log_path = config.log + opt.log + '/'
if not os.path.exists(log_path):
os.mkdir(log_path)
logging = utils.logging(
log_path + 'log.txt') # 这种方式也值得学习，单独写一个logging的函数，直接调用，既print，又记录到Log文件里。
logging_csv = utils.logging_csv(log_path + 'record.csv')
for k, v in config.items():
logging("%s:\t%s\n" % (str(k), str(v)))
logging("\n")
logging(repr(model) + "\n\n")
logging('total number of parameters: %d\n\n' % param_count)
logging('score function is %s\n\n' % opt.score)
if opt.restore:
updates = checkpoints['updates']
else:
updates = 0
total_loss, start_time = 0, time.time()
report_total, report_correct = 0, 0
else:
updates = 0
# ======================================================================================================================
"""log config"""
if not os.path.exists(config.log):
os.mkdir(config.log)
if opt.log == '':
log_path = config.log + utils.format_time(time.localtime()) + '/'
else:
log_path = config.log + opt.log + '/'
if not os.path.exists(log_path):
os.mkdir(log_path)
logging = utils.logging(log_path+'model_config.txt')
logging_csv = utils.logging_csv(log_path + 'model_record.csv')
for k, v in config.items():
logging("%s:\t%s\n" % (str(k), str(v)))
logging("\n")
logging(repr(model)+"\n\n")
logging('total number of parameters: %d\n\n' % param_count)
# ======================================================================================================================
"""train"""
scores = [[] for metric in config.metric]
scores = collections.OrderedDict(zip(config.metric, scores))
loss_function = nn.CrossEntropyLoss()
def train(epoch):
for param in model.parameters():
param_count += param.view(-1).size()[0]
if not os.path.exists(config.log):
os.makedirs(config.log)
if config.log.endswith('/'):
log_path = config.log
else:
log_path = config.log + '/'
if not os.path.exists(log_path):
os.mkdir(log_path)
logging = utils.logging(log_path + 'log.txt')
if not opt.notrain:
logging_train_loss = utils.logging_csv(log_path + 'train_loss.csv',
['epoch', 'updates', 'log_loss'])
logging_valid_loss = utils.logging_csv(log_path + 'valid_loss.csv',
['epoch', 'updates', 'log_loss'])
logging_metric = utils.logging_dict_csv(log_path + 'metrics.csv',
['epoch', 'updates'] + all_metrics)
for k, v in config.items():
logging("%s:\t%s\n" % (str(k), str(v)))
logging("\n")
logging(repr(model) + "\n\n")
logging('total number of parameters: %d\n\n' % param_count)
if opt.restore:
updates = checkpoints['updates']
else:
for param in model.parameters():
param_count += param.view(-1).size()[0]
# log为记录文件
# config.log是记录的文件夹, 最后一定是/
# opt.log是此次运行时记录的文件夹的名字
if not os.path.exists(config.log):
os.mkdir(config.log)
if opt.log == '':
log_path = config.log + utils.format_time(time.localtime()) + '/'
else:
log_path = config.log + opt.log + '/'
if not os.path.exists(log_path):
os.mkdir(log_path)
logging = utils.logging(log_path+'log.txt') # 往这个文件里写记录
logging_csv = utils.logging_csv(log_path+'record.csv') # 往这个文件里写记录
for k, v in config.items():
logging("%s:\t%s\n" % (str(k), str(v)))
logging("\n")
logging(repr(model)+"\n\n") # 记录这个文件的框架
logging('total number of parameters: %d\n\n' % param_count)
logging('score function is %s\n\n' % opt.score)
# updates是已经进行了几个epoch, 防止中间出现程序中断的情况.
if opt.restore:
updates = checkpoints['updates']
else:
updates = 0
total_loss, start_time = 0, time.time()