QANet_dureader icon indicating copy to clipboard operation
QANet_dureader copied to clipboard

关于loss变得异常大的问题

Open tomtang110 opened this issue 6 years ago • 11 comments

2019-01-19 17 12 46 loss在某一部分数据变得异常大,最后导致验证集所有分数为0. 请问,是怎么产生的呢?

tomtang110 avatar Jan 19 '19 09:01 tomtang110

您好,我也遇到了,这类情况,请问您解决了吗,可以分享下吗

pluto-junzeng avatar Mar 25 '19 12:03 pluto-junzeng

total_loss += loss * len(batch['raw_data'])

pluto-junzeng avatar Mar 25 '19 12:03 pluto-junzeng

换模型吧,QA模型对中文数据集敏感太大。

获取 Outlook for Androidhttps://aka.ms/ghei36


From: 曾俊 [email protected] Sent: Monday, March 25, 2019 11:01:06 PM To: SeanLee97/QANet_dureader Cc: tomtang110; Author Subject: Re: [SeanLee97/QANet_dureader] 关于loss变得异常大的问题 (#8)

您好,我也遇到了,这类情况,请问您解决了吗,可以分享下吗

― You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/SeanLee97/QANet_dureader/issues/8#issuecomment-476166454, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AfwcQhhPHWl_cyxYiKoXlDaofMye6gKzks5vaLqCgaJpZM4aJC_z.

tomtang110 avatar Mar 27 '19 22:03 tomtang110

我也遇到了这种情况

YeDeming avatar May 31 '19 07:05 YeDeming

拿全部数据集训练也有这个问题,确定代码没问题??

fanweifengdw avatar Jun 06 '19 03:06 fanweifengdw

检验过代码确实没啥明显问题

获取 Outlook for Androidhttps://aka.ms/ghei36


From: fanweifengdw [email protected] Sent: Thursday, June 6, 2019 1:06:33 PM To: SeanLee97/QANet_dureader Cc: tomtang110; Author Subject: Re: [SeanLee97/QANet_dureader] 关于loss变得异常大的问题 (#8)

拿全部数据集训练也有这个问题,确定代码没问题??

― You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/SeanLee97/QANet_dureader/issues/8?email_source=notifications&email_token=AH6BYQUANL4IPVGATA4CXS3PZB5LTA5CNFSM4GREF7Z2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODXBS6XI#issuecomment-499330909, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AH6BYQQS6ISW2GXGB4QUSTTPZB5LTANCNFSM4GREF7ZQ.

tomtang110 avatar Jun 06 '19 03:06 tomtang110

有没有报错的代码吗?我直接用代码报TypeError: Failed to convert object of type <class 'list'> to Tensor. Contents: [32000, 20, Dimension(32)]. Consider casting elements to a supported type.错误

sarahyq avatar Jun 20 '19 09:06 sarahyq

这个真不适合中文,都什么年代了用bert吧,可以训练但是中文数据集不好

获取 Outlook for Androidhttps://aka.ms/ghei36


From: sarahyq [email protected] Sent: Thursday, June 20, 2019 7:42:10 PM To: SeanLee97/QANet_dureader Cc: tomtang110; Author Subject: Re: [SeanLee97/QANet_dureader] 关于loss变得异常大的问题 (#8)

有没有报错的代码吗?我直接用代码报TypeError: Failed to convert object of type <class 'list'> to Tensor. Contents: [32000, 20, Dimension(32)]. Consider casting elements to a supported type.错误

― You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/SeanLee97/QANet_dureader/issues/8?email_source=notifications&email_token=AH6BYQVNIAXMOPTPP7NAFYLP3NGHFA5CNFSM4GREF7Z2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODYE4OMY#issuecomment-503957299, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AH6BYQQX6M7KM5Q2NGABWELP3NGHFANCNFSM4GREF7ZQ.

tomtang110 avatar Jun 20 '19 10:06 tomtang110

换模型吧,QA模型对中文数据集敏感太大。 获取 Outlook for Androidhttps://aka.ms/ghei36 ________________________________ From: 曾俊 [email protected] Sent: Monday, March 25, 2019 11:01:06 PM To: SeanLee97/QANet_dureader Cc: tomtang110; Author Subject: Re: [SeanLee97/QANet_dureader] 关于loss变得异常大的问题 (#8) 您好,我也遇到了,这类情况,请问您解决了吗,可以分享下吗 ― You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#8 (comment)>, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AfwcQhhPHWl_cyxYiKoXlDaofMye6gKzks5vaLqCgaJpZM4aJC_z. 有分析出为什么中文不好吗?

sarahyq avatar Jun 20 '19 10:06 sarahyq

这种情况非常罕见,我觉得可能涉及到,语义的一些东西在里面,只能说QAnet这个网络设计的不够general。你看同是selfattention机制,bert和今天才出的xlnet就具有泛化。

获取 Outlook for Androidhttps://aka.ms/ghei36


From: sarahyq [email protected] Sent: Thursday, June 20, 2019 8:26:32 PM To: SeanLee97/QANet_dureader Cc: tomtang110; Author Subject: Re: [SeanLee97/QANet_dureader] 关于loss变得异常大的问题 (#8)

换模型吧,QA模型对中文数据集敏感太大。 获取 Outlook for Androidhttps://aka.ms/ghei36 … ________________________________ From: 曾俊 [email protected]mailto:[email protected] Sent: Monday, March 25, 2019 11:01:06 PM To: SeanLee97/QANet_dureader Cc: tomtang110; Author Subject: Re: [SeanLee97/QANet_dureader] 关于loss变得异常大的问题 (#8https://github.com/SeanLee97/QANet_dureader/issues/8) 您好,我也遇到了,这类情况,请问您解决了吗,可以分享下吗 D You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#8 (comment)https://github.com/SeanLee97/QANet_dureader/issues/8#issuecomment-476166454>, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AfwcQhhPHWl_cyxYiKoXlDaofMye6gKzks5vaLqCgaJpZM4aJC_z. 有分析出为什么中文不好吗?

― You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/SeanLee97/QANet_dureader/issues/8?email_source=notifications&email_token=AH6BYQXBMMROYHGXXABG6MLP3NLNRA5CNFSM4GREF7Z2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODYFAAMI#issuecomment-503971889, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AH6BYQU6GFB763LRYERUUKTP3NLNRANCNFSM4GREF7ZQ.

tomtang110 avatar Jun 20 '19 10:06 tomtang110

The model can be used for production?

weinixuehao avatar Sep 03 '19 07:09 weinixuehao