李炜

attn可以直接删去，这个是分析的时候看attention 权重的发自我的华为手机-------- 原始邮件 --------主题：Re: [lancopku/Graph-to-seq-comment-generation] run error (#7)发件人：shenjian5 收件人：lancopku/Graph-to-seq-comment-generation 抄送：Subscribed Traceback (most recent call last): File "train.py", line 299, in main() File "train.py", line 291, in main max_bleu = train(model,...

you can use the word with top 60000 frequency as the vocab

Sorry for the inconvenience, going to complete the readme file soon.

No, they are not used in our method发自我的荣耀手机-------- 原始邮件 --------发件人： MollyShuu 日期： 2021年2月27日周六 傍晚6:52收件人： lancopku/Graph-to-seq-comment-generation 抄送： Subscribed 主 题： [lancopku/Graph-to-seq-comment-generation] About Concept Community mining in ccig.py file (#12) Hello, I...

There is no extra graph libraries needed

药物的embedding是用word2vec预训练的

我们关注的是F值，precision确实会有些下降