exit(1208)
print_info('{}init pram{}'.format('*' * 15, '*' * 15))
loss = LossInForward()
encoding_type = 'bmeso'
f1_metric = SpanFPreRecMetric(vocabs['label'],
pred='pred',
target='target',
seq_len='seq_len',
encoding_type=encoding_type)
acc_metric = AccuracyMetric(
pred='pred',
target='target',
seq_len='seq_len',
)
acc_metric.set_metric_name('label_acc')
metrics = [f1_metric, acc_metric]
if args.self_supervised:
chars_acc_metric = AccuracyMetric(pred='chars_pred',
target='chars_target',
seq_len='seq_len')
chars_acc_metric.set_metric_name('chars_acc')
metrics.append(chars_acc_metric)
if args.see_param:
for n, p in model.named_parameters():
print_info('{}:{}'.format(n, p.size()))
print_info('see_param mode: finish')
if not args.debug:
exit(1208)
datasets['train'].apply
elif args.init == 'norm':
logging.info('xavier norm init : {}'.format(n))
nn.init.xavier_normal_(p)
except:
logging.info(n)
exit(1208)
logging.info('{}init pram{}\n'.format('*' * 15, '*' * 15))
loss = LossInForward()
f1_metric = SpanFPreRecMetric(vocabs['label'],
pred='pred',
target='target',
seq_len='seq_len',
encoding_type='bio')
acc_metric = AccuracyMetric(pred='pred', target='target', seq_len='seq_len')
acc_metric.set_metric_name('label_acc')
metrics = [f1_metric, acc_metric]
# if args.self_supervised:
# chars_acc_metric = AccuracyMetric(pred='chars_pred', target='chars_target', seq_len='seq_len')
# chars_acc_metric.set_metric_name('chars_acc')
# metrics.append(chars_acc_metric)
embedding_param = list(model.bigram_embed.parameters()) + list(
model.lattice_embed.parameters())
embedding_param_ids = list(map(id, embedding_param))
non_embedding_param = list(
filter(lambda x: id(x) not in embedding_param_ids, model.parameters()))
param_ = [{
'params': non_embedding_param