初学者求lstm+cnn/tensorflow代码?|New beginner for "lstm+cnn/tensorflow"
Opened this issue · 8 comments
WaveLi123 commented
初学者求lstm+cnn/tensorflow代码?|New beginner for "lstm+cnn/tensorflow"
pcgreat commented
chenzhi1992 commented
tf版本的lstm+cnn的代码,怎么还是cnn的代码啊?@ pcgreat
pcgreat commented
@chenzhi1992 抱歉,上传代码的时候弄错了,现在更正了
chenzhi1992 commented
谢谢回复!
发自网易邮箱大师
在2017年07月21日 23:55,Chao Pan 写道:
@chenzhi1992 抱歉,上传代码的时候弄错了,现在更正了
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub, or mute the thread.
chenzhi1992 commented
你的lstm+cnn 大概训练多少step后loss收敛?我训练到1000step,acc一直在0.1到0.2之间.@pcgreat
@ pcgreat
pcgreat commented
1000步太少了,至少12k步才能看出点效果。你可以适当加一点优化,比如有选择地抽取negative的sample,这样收敛快一些
…Sent from my iPhone
On Jul 23, 2017, at 9:21 PM, chenzhi1992 ***@***.***> wrote:
你的lstm+cnn ***@***.***
@ pcgreat
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub, or mute the thread.
zifengwen commented
cnn/tensorflow/insqa_cnn 中:
with tf.name_scope("accuracy"):
self.correct = tf.equal(zero, self.losses)
self.accuracy = tf.reduce_mean(tf.cast(self.correct, "float"), name="accuracy")
这几行不足以评估模型的好坏吧,只是统计训练数据中“一个正确答案跟一个错误答案”选对正确答案的比例。。
zifengwen commented
cnn/tensorflow/insqa_train 中,模型评估的方法dev_step()中,实际上只评估了20条数据吧。
为什么不是直接所有答案的向量计算出来,,然后对于新的问题,计算向量,匹配所有答案向量,得到最佳答案,跟真实的答案比较得到准确率呢??
现在这样写有什么好处??