Ziqiang Liu

> ### [](/Henry-soul) ** [Henry](/Henry-soul)** 你好，我只跑通了proto，并未提交测试

> Hi, thanks for your contribution. However,when i work with this code , i found there are two files i dont know how to obtain. Despite i checked the link...

Thanks for replying. I have only got "glove.6B.50d.txt", but I didn't have "glove.6B.50d_word2id.json" and "glove.6B.50d_mat.npy" these two files, how can I get these or how to generate them? Look forward to...

Thank you so much! ------------------ 原始邮件 ------------------ 发件人: "rtmaww/SENT" ***@***.***>; 发送时间: 2021年11月24日(星期三) 晚上8:45 ***@***.***>; ***@***.******@***.***>; 主题: Re: [rtmaww/SENT] How could i get "glove.6B.50d_word2id.json" and "glove.6B.50d_mat.npy"? (Issue #9) Hello. Thank you for reminding me this...

> **Problem or Use Case** I had a bug in my script that `wandb.log` very large array. wandb hanged at `wandb: - 0.00MB of 0.09MB uploaded (0.00MB deduped)` and timed...

> 您具体可以仔细看看论文数据集的划分以及模型介绍： Three parts: training set which consists of 54 relations (i.e., base relations R_base) each with 550 instances; validation set which consists of 54 relations (i.e., base relations R_base)...

> > > 您具体可以仔细看看论文数据集的划分以及模型介绍： Three parts: training set which consists of 54 relations (i.e., base relations R_base) each with 550 instances; validation set which consists of 54 relations (i.e., base...

> > 您好，我尝试了很多次都无法复现出论文的结果，请问可不可以提供一下能够复现出论文结果的参数？感激不尽！ > > 参数列表说明： > > 1. 学习率 默认：2e-3 > 2. 第一阶段模型预训练的Base模型：采用可视化或调整迭代次数超参或采用验证集验证等方式选择最佳的预训练Model; 默认取：7000 iter > 3. margin参数：该参数指的是计算triplet loss时，负样本与正样本的距离；由于无法预先估计多维特征空间上的正负例之间距离，导致该参数对模型的稳定性影响较大；调整范围：(0.0, 30.0), 默认设置：3.0； > 4. triplet_num: 调整范围：(1, 20)； 默认：20 > 5....

> > > 你好，我这边复现的下论文结果，但是发现，BERT的1shot和5shot，跑不出论文的效果，请问是参数不对么，还是其他原因呢？（其他都可以） > > > 我跑的是： > > > 1 shot > > > base | novel | both > > > 80.12 | 60.06| 78.42 > >...

> Hi all, thanks for your patience, we'll respond more promptly going forwards. > > The issues mentioned here look quite reminiscent of some of the failure modes we've been...