提交 3f3caf6e 编写于 作者: B breezedeus

fix: _loss_type should be set to None for inference

上级 d2bfade6
......@@ -10,6 +10,7 @@ class CnHyperparams(object):
# self._train_epoch_size = 2560000
# self._eval_epoch_size = 3000
self._num_epoch = 20
self._loss_type = 'ctc'
self.optimizer = "Adam"
self._learning_rate = 0.001
self.wd = 0.00001
......@@ -71,7 +72,7 @@ class CnHyperparams(object):
@property
def loss_type(self):
return "ctc"
return self._loss_type
@property
def batch_size(self):
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册