yuan2961634811

Results 5 comments of yuan2961634811

你好,从统计角度讲,对结果没有影响。当然您也可以去除重复的都可以。 在 2021-01-04 10:41:06,"Kaibing Chen" 写道: 是的 def random_negs(l,r,no,s): # set_s=set(s) negs = [] for i in range(no): t = np.random.randint(l, r) # while (t in set_s): while (t== s):...

你好,只需要所有baseline的evaluation保持一致就可以哈 在 2021-01-04 10:50:40,"Kaibing Chen" 写道: 你好,从统计角度讲,对结果没有影响。当然您也可以去除重复的都可以。 对于年龄性别这些下游任务,类别都较少,随机采负样本测试,统计上指标都会偏高的,整体上相当于测试的类别变少了 实际跑的时候,不去重跑出来与你论文的指标差不多,稍微低点;去重后指标要低个10个点左右 — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.

[1, 2, 4, 8, 1, 2, 4, 8, 1, 2, 4, 8, 1, 2, 4, 8] 是32层,1 4 means 1 2 4 8. 一般2 iterations 就收敛了, 不用400的 (请阅读readme里面哈,400不是默认设置). 你用NextitNet_TF_Pretrain_topk.py .py试一下。另外...

Hi it has no impact, you can also mask zero padding which is faster but the accuracy should be more or less the same