能否提供一个论文里的FastReID-R50的config文件？
qianjinhao opened this issue · 9 comments
作者你好，非常感谢你的工作，我想要自己训练个比较高的baseline，但是使用fastreid原本的config结果差的比较远，能否提供一个论文里的FastReID-R50的config文件？我觉得可能在某些超参设置或者数据增强方面存在差异。谢谢！
请问你是跑什么setting 发件人: qianjinhao @.> 答复: xiaomingzhid/sskd @.> 日期: 2021年10月29日 星期五 上午11:04 收件人: xiaomingzhid/sskd @.> 抄送: Subscribed @.> 主题: [来自外部的邮件][xiaomingzhid/sskd] 能否提供一个论文里的FastReID-R50的config文件？ (Issue #3) 作者你好，非常感谢你的工作，我想要自己训练个比较高的baseline，但是使用fastreid原本的config结果差的比较远，能否提供一个论文里的FastReID-R50的config文件？我觉得可能在某些超参设置或者数据增强方面存在差异。谢谢！ — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub<#3>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AVCF6EKOZ3M43YN6AQ7LRJLUJIFM5ANCNFSM5G6JXGGQ. Triage notifications on the go with GitHub Mobile for iOShttps://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Androidhttps://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.
我用了r50 bot 跑的M+D+C->MS，没有修改其他，rank-1 28.86 mAP 10.73
其余的参数配置不变，改为adam优化器及其对应的学习率就行 发件人: qianjinhao @.> 答复: xiaomingzhid/sskd @.> 日期: 2021年10月29日 星期五 上午11:23 收件人: xiaomingzhid/sskd @.> 抄送: 何凌霄 @.>, Comment @.***> 主题: [来自外部的邮件]Re: [xiaomingzhid/sskd] 能否提供一个论文里的FastReID-R50的config文件？ (Issue #3) 请问你是跑什么setting 发件人: qianjinhao @.> 答复: xiaomingzhid/sskd @.> 日期: 2021年10月29日 星期五 上午11:04 收件人: xiaomingzhid/sskd @.> 抄送: Subscribed @.> 主题: [来自外部的邮件][xiaomingzhid/sskd] 能否提供一个论文里的FastReID-R50的config文件？ (Issue #3<#3>) 作者你好，非常感谢你的工作，我想要自己训练个比较高的baseline，但是使用fastreid原本的config结果差的比较远，能否提供一个论文里的FastReID-R50的config文件？我觉得可能在某些超参设置或者数据增强方面存在差异。谢谢！ — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub<#3<#3>>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AVCF6EKOZ3M43YN6AQ7LRJLUJIFM5ANCNFSM5G6JXGGQ. Triage notifications on the go with GitHub Mobile for iOShttps://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Androidhttps://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub. 我用了r50 bot 跑的M+D+C->MS，没有修改其他，rank-1 28.86 mAP 10.73 — You are receiving this because you commented. Reply to this email directly, view it on GitHub<#3 (comment)>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AVCF6ENY5FLAOKCBN6IO3EDUJIHRDANCNFSM5G6JXGGQ. Triage notifications on the go with GitHub Mobile for iOShttps://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Androidhttps://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub
是bot、agw还是sbs呢？你说的学习率指的是根据batchsize修改吗？
Batchsize不变 ， OPT:Adam MAX_EPOCH: 30 BASE_LR: 0.00035 BIAS_LR_FACTOR: 1. WEIGHT_DECAY: 0.0005 WEIGHT_DECAY_BIAS: 0.0005 IMS_PER_BATCH: 64 SCHED: CosineAnnealingLR DELAY_EPOCHS: 30 ETA_MIN_LR: 0.00000077 WARMUP_FACTOR: 0.1 WARMUP_ITERS: 500 FREEZE_ITERS: 500 CHECKPOINT_PERIOD: 10 发件人: qianjinhao @.> 答复: xiaomingzhid/sskd @.> 日期: 2021年10月29日 星期五 上午11:28 收件人: xiaomingzhid/sskd @.> 抄送: 何凌霄 @.>, Comment @.> 主题: [来自外部的邮件]Re: [xiaomingzhid/sskd] 能否提供一个论文里的FastReID-R50的config文件？ (Issue #3) 其余的参数配置不变，改为adam优化器及其对应的学习率就行 发件人: qianjinhao @.> 答复: xiaomingzhid/sskd @.> 日期: 2021年10月29日 星期五 上午11:23 收件人: xiaomingzhid/sskd @.> 抄送: 何凌霄 @.>, Comment @.> 主题: [来自外部的邮件]Re: [xiaomingzhid/sskd] 能否提供一个论文里的FastReID-R50的config文件？ (Issue #3<#3>) 请问你是跑什么setting 发件人: qianjinhao @.> 答复: xiaomingzhid/sskd @.> 日期: 2021年10月29日 星期五 上午11:04 收件人: xiaomingzhid/sskd @.> 抄送: Subscribed @.> 主题: [来自外部的邮件][xiaomingzhid/sskd] 能否提供一个论文里的FastReID-R50的config文件？ (Issue #3<#3><#3#3>) 作者你好，非常感谢你的工作，我想要自己训练个比较高的baseline，但是使用fastreid原本的config结果差的比较远，能否提供一个论文里的FastReID-R50的config文件？我觉得可能在某些超参设置或者数据增强方面存在差异。谢谢！ — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub<#3<#3><#3#3>>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AVCF6EKOZ3M43YN6AQ7LRJLUJIFM5ANCNFSM5G6JXGGQ. Triage notifications on the go with GitHub Mobile for iOShttps://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Androidhttps://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub. 我用了r50 bot 跑的M+D+C->MS，没有修改其他，rank-1 28.86 mAP 10.73 — You are receiving this because you commented. Reply to this email directly, view it on GitHub<#3 (comment)<#3 (comment)>>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AVCF6ENY5FLAOKCBN6IO3EDUJIHRDANCNFSM5G6JXGGQ. Triage notifications on the go with GitHub Mobile for iOShttps://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Androidhttps://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub 是bot、agw还是sbs呢？你说的学习率指的是根据batchsize修改吗？ — You are receiving this because you commented. Reply to this email directly, view it on GitHub<#3 (comment)>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AVCF6EPJL22GZSU7XFK5YOTUJIIEPANCNFSM5G6JXGGQ. Triage notifications on the go with GitHub Mobile for iOShttps://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Androidhttps://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.
请教一下，大的batchsize会不会导致结果变差？
即使根据作者的修改还是跟论文里的结果有一定的距离？