lijianlong1 / data_uhnd_other

2 stars 1 forks source link

在复现实验中出现一些问题,如果可以解答非常感谢 #1

Open Joshuass11 opened 6 months ago

Joshuass11 commented 6 months ago

dl_module = globalpointer_NER(r'C:\Users\ASUS\Desktop\data_uhnd_other-master\chinese-roberta-wwm', n_head=len(label_list)) 在这一行里,需要填入的是目录下的json文件?这个实验的fastNLP版本是多少呢?需要用到tensorflow框架吗

lijianlong1 commented 6 months ago

可以看我比赛那个git,写的比较细致,ccl比赛的

---原始邮件--- 发件人: @.> 发送时间: 2024年3月7日(周四) 凌晨3:32 收件人: @.>; 抄送: @.***>; 主题: [lijianlong1/data_uhnd_other] 在复现实验中出现一些问题,如果可以解答非常感谢 (Issue #1)

nezha_global_fgm_train.py应该是唯一的主函数?在复现本文实验时,应该调用的是model.py和nerpre.py两个文件,对吗? 在调用路径时以下路径中只包含一个txt文件,这里模型要求的超参数应该在哪里取得呢? '/opt/lijianlong/gaiic/chinese-roberta-wwm' globalpointer_NER('/opt/lijianlong/gaiic/chinese-roberta-wwm', n_head=len(label_list))

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.Message ID: @.***>

Joshuass11 commented 6 months ago

好的,感谢分享

---原始邮件--- 发件人: @.> 发送时间: 2024年3月7日(周四) 晚上6:51 收件人: @.>; 抄送: @.**@.>; 主题: Re: [lijianlong1/data_uhnd_other] 在复现实验中出现一些问题,如果可以解答非常感谢 (Issue #1)

可以看我比赛那个git,写的比较细致,ccl比赛的

---原始邮件--- 发件人: @.> 发送时间: 2024年3月7日(周四) 凌晨3:32 收件人: @.>; 抄送: @.***>; 主题: [lijianlong1/data_uhnd_other] 在复现实验中出现一些问题,如果可以解答非常感谢 (Issue #1)

nezha_global_fgm_train.py应该是唯一的主函数?在复现本文实验时,应该调用的是model.py和nerpre.py两个文件,对吗? 在调用路径时以下路径中只包含一个txt文件,这里模型要求的超参数应该在哪里取得呢? '/opt/lijianlong/gaiic/chinese-roberta-wwm' globalpointer_NER('/opt/lijianlong/gaiic/chinese-roberta-wwm', n_head=len(label_list))

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.Message ID: @.> — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.>

Joshuass11 commented 5 months ago

你好!我是之前在github上提问的人,在尝试复现您的实验(FGM对抗学习)。方便加下微信吗?有一些代码细节的问题想请教下。打扰啦,感谢回答之前的问题

---原始邮件--- 发件人: @.> 发送时间: 2024年3月7日(周四) 晚上6:51 收件人: @.>; 抄送: @.**@.>; 主题: Re: [lijianlong1/data_uhnd_other] 在复现实验中出现一些问题,如果可以解答非常感谢 (Issue #1)

可以看我比赛那个git,写的比较细致,ccl比赛的

---原始邮件--- 发件人: @.> 发送时间: 2024年3月7日(周四) 凌晨3:32 收件人: @.>; 抄送: @.***>; 主题: [lijianlong1/data_uhnd_other] 在复现实验中出现一些问题,如果可以解答非常感谢 (Issue #1)

nezha_global_fgm_train.py应该是唯一的主函数?在复现本文实验时,应该调用的是model.py和nerpre.py两个文件,对吗? 在调用路径时以下路径中只包含一个txt文件,这里模型要求的超参数应该在哪里取得呢? '/opt/lijianlong/gaiic/chinese-roberta-wwm' globalpointer_NER('/opt/lijianlong/gaiic/chinese-roberta-wwm', n_head=len(label_list))

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.Message ID: @.> — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.>

lijianlong1 commented 5 months ago

可以看一下,具体的fgm扰动的代码实现例子方法,这块是为了最大化扰动,挨个进行回传,而不是相加回传

---原始邮件--- 发件人: @.> 发送时间: 2024年4月12日(周五) 下午4:23 收件人: @.>; 抄送: @.**@.>; 主题: Re: [lijianlong1/data_uhnd_other] 在复现实验中出现一些问题,如果可以解答非常感谢 (Issue #1)

你好!我是之前在github上提问的人,在尝试复现您的实验(FGM对抗学习)。方便加下微信吗?有一些代码细节的问题想请教下。打扰啦,感谢回答之前的问题

---原始邮件--- 发件人: @.> 发送时间: 2024年3月7日(周四) 晚上6:51 收件人: @.>; 抄送: @.**@.>; 主题: Re: [lijianlong1/data_uhnd_other] 在复现实验中出现一些问题,如果可以解答非常感谢 (Issue #1)

可以看我比赛那个git,写的比较细致,ccl比赛的

---原始邮件---
发件人: @.>
发送时间: 2024年3月7日(周四) 凌晨3:32
收件人:
@.>;
抄送: @.***>;
主题: [lijianlong1/data_uhnd_other] 在复现实验中出现一些问题,如果可以解答非常感谢 (Issue #1)

nezha_global_fgm_train.py应该是唯一的主函数?在复现本文实验时,应该调用的是model.py和nerpre.py两个文件,对吗?
在调用路径时以下路径中只包含一个txt文件,这里模型要求的超参数应该在哪里取得呢?
'/opt/lijianlong/gaiic/chinese-roberta-wwm'
globalpointer_NER('/opt/lijianlong/gaiic/chinese-roberta-wwm', n_head=len(label_list))


Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.Message ID: @.>
— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID:
@.> — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>

lijianlong1 commented 5 months ago

微信,18801275087

---原始邮件--- 发件人: @.> 发送时间: 2024年4月12日(周五) 下午4:23 收件人: @.>; 抄送: @.**@.>; 主题: Re: [lijianlong1/data_uhnd_other] 在复现实验中出现一些问题,如果可以解答非常感谢 (Issue #1)

你好!我是之前在github上提问的人,在尝试复现您的实验(FGM对抗学习)。方便加下微信吗?有一些代码细节的问题想请教下。打扰啦,感谢回答之前的问题

---原始邮件--- 发件人: @.> 发送时间: 2024年3月7日(周四) 晚上6:51 收件人: @.>; 抄送: @.**@.>; 主题: Re: [lijianlong1/data_uhnd_other] 在复现实验中出现一些问题,如果可以解答非常感谢 (Issue #1)

可以看我比赛那个git,写的比较细致,ccl比赛的

---原始邮件---
发件人: @.>
发送时间: 2024年3月7日(周四) 凌晨3:32
收件人:
@.>;
抄送: @.***>;
主题: [lijianlong1/data_uhnd_other] 在复现实验中出现一些问题,如果可以解答非常感谢 (Issue #1)

nezha_global_fgm_train.py应该是唯一的主函数?在复现本文实验时,应该调用的是model.py和nerpre.py两个文件,对吗?
在调用路径时以下路径中只包含一个txt文件,这里模型要求的超参数应该在哪里取得呢?
'/opt/lijianlong/gaiic/chinese-roberta-wwm'
globalpointer_NER('/opt/lijianlong/gaiic/chinese-roberta-wwm', n_head=len(label_list))


Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.Message ID: @.>
— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID:
@.> — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>