training time and inference time

[Lily-Git-hub]

Hi all,

Thank you for sharing your great work. Could you please share the training and inference times for the different models in your paper (e.g., VAE, diffusion model, appearance model)?

Thank you!
Best regards,
lily

[KickCellarDoor]

Training time is listed in the paper. However, training GSM takes much longer for me compared to what the author reported. (GSM: paper, 8xa100x2d; me, 8xa100x5d )

[chris-aeviator]

Training time is listed in the paper. However, training GSM takes much longer for me compared to what the author reported. (GSM: paper, 8xa100x2d; me, 8xa100x5d )

@KickCellarDoor would it be possible for you to share the trained checkpoint?

[ChnWuyue]

论文中列出了培训时间。然而，与作者报告的相比，训练 GSM 对我来说需要更长的时间。（GSM：纸质，8xa100x2d;me，8xa100x5d）

@KickCellarDoor would it be possible for you to share the trained checkpoint?

[KickCellarDoor]

论文中列出了培训时间。然而，与作者报告的相比，训练 GSM 对我来说需要更长的时间。（GSM：纸质，8xa100x2d;me，8xa100x5d）

@KickCellarDoor would it be possible for you to share the trained checkpoint?

Sorry, the checkpoints won't be available.

[ChnWuyue]

论文中列出了培训时间。然而，与作者报告的相比，训练 GSM 对我来说需要更长的时间。（GSM：纸质，8xa100x2d;me，8xa100x5d）

@KickCellarDoor would it be possible for you to share the trained checkpoint?

Sorry, the checkpoints won't be available.

Could you kindly share whether the reconstruction results match those reported in the paper? Did you modify any of the official config files? I haven’t been able to reproduce them on my side—any insights would be really appreciated.

[KickCellarDoor]

My results looks similar to what they reported. Best regards, Xingxin

…

________________________________ 发件人: 伍h ***@***.***> 发送时间: Friday, September 12, 2025 3:14:47 PM 收件人: nv-tlabs/SCube ***@***.***> 抄送: Eric Chen ***@***.***>; Mention ***@***.***> 主题: Re: [nv-tlabs/SCube] training time and inference time (Issue #9) [https://avatars.githubusercontent.com/u/58622809?s=20&v=4]ChnWuyue left a comment (nv-tlabs/SCube#9)<#9 (comment)> 论文中列出了培训时间。然而，与作者报告的相比，训练 GSM 对我来说需要更长的时间。（GSM：纸质，8xa100x2d;me，8xa100x5d） @KickCellarDoor<https://github.com/KickCellarDoor> would it be possible for you to share the trained checkpoint? Sorry, the checkpoints won't be available. Could you kindly share whether the reconstruction results match those reported in the paper? Did you modify any of the official config files? I haven’t been able to reproduce them on my side―any insights would be really appreciated. ― Reply to this email directly, view it on GitHub<#9 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ADURMTMSXGAV4ASYLW7XKSL3SJXGPAVCNFSM6AAAAABVLDG6EGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTEOBUGA3DAMBQHA>. You are receiving this because you were mentioned.Message ID: ***@***.***>

[ChnWuyue]

My results looks similar to what they reported.

Best regards,
Xingxin
…

thanks for your reply, did you modify any of the official config files. it's important for me.