# get the pretrained model prefix
if params.load_model is None:
pretrained_prefix = ''
else:
pretrained_prefix = os.path.basename(params.load_model)
pretrained_prefix = '_' + pretrained_prefix[:pretrained_prefix.
find('.pth')]
mlp_str = params.cldc_mlp_type if params.cldc_mlp_type == 'sl' else (
params.cldc_mlp_type + params.activation +
str(params.cldc_do_rate))
cldc_str = (
'{}red.{}_'.format(params.task_type, params.cldc_train_mode) +
('{}_'.format(params.semicldc_yprior_type)
if params.task_type == 'semi-cldc' else '') +
('{}_'.format(params.semicldc_cond_type)
if params.task_type == 'semi-cldc' else '') +
'train.{}_'.format(params.cldc_train_scale) +
'val.{}_'.format(params.CLDC_VAL_EVERY) +
'{}_'.format('2'.join(params.cldc_part)) + '{}_'.format(mlp_str) +
'lr.{}_'.format(params.cldc_init_learning_rate) +
'per.{}_'.format(params.cldc_percentile) + 'pre.{}'.format(
'n' if params.pretrained is None else params.pretrained))
out_str = cldc_str + pretrained_prefix + '_' + cur_time
return out_str
if __name__ == '__main__':
arguments = create_args()
params = Params(arguments)
main(params)
self.wdidx2moidx = torch.LongTensor(self.wdidx2moidx)
self.momask = torch.Tensor(self.momask)
if compf.startswith('pp') or compf.startswith('mp'):
self.wdidx2pos = torch.LongTensor(self.wdidx2pos)
if __name__ == '__main__':
import sys
import os
sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
import torch
from args import create_args
from sms.sms_input_data import SmsInputData
from morf.morf_input_data import MorfInputData
from bpe.bpe_input_data import BPEInputData
from charn.charn_input_data import CharnInputData
args = create_args()
if args.subword == 'sms':
data = SmsInputData(args)
elif args.subword == 'morf':
data = MorfInputData(args)
elif args.subword == 'bpe':
data = BPEInputData(args)
elif args.subword == 'charn':
data = CharnInputData(args)
state_dict = {'mo2idx': data.mo2idx,
'word2idx': data.word2idx}
torch.save(state_dict, args.save_model)
# Progress bar
if opt.local_rank == 0:
pb.set_postfix(loss=loss,
disto=disto,
rate_y_0=rate_y_0,
rate_z=rate_z,
refresh=False)
pb.update()
except KeyboardInterrupt:
status_code = 130
if opt.local_rank == 0:
pb.close()
print('Done')
# Save model
print('Saving...')
torch.save(model.state_dict(), os.path.join(opt.save_path, 'model.pt'))
return status_code
if __name__ == '__main__':
# Arguments
p = args.create_args()
# Parse arguments
opt = helper.DotDict(vars(p.parse_args()))
# Disable output for all processes but the first one
if opt.local_rank != 0:
sys.stdout = open(os.devnull, "w")
# Main
main(opt)