StepFormer
StepFormer copied to clipboard
SamsungLabs
Reproduce paper's results
Dear authors,
Thanks for the great work and for sharing the code!
I want to reproduce the results you had in the paper. Are the checkpoints accessible? Regarding the datasets, can you share the download links or point me to the correct repo?
I appreciate any help you can provide.
Kind regards, Alessandro Flaborea
Updated the saved model in the repo. https://github.com/SamsungLabs/StepFormer/blob/main/packed_model.zip
Updated the saved model in the repo. https://github.com/SamsungLabs/StepFormer/blob/main/packed_model.zip
Hi, when I use git lfs pull to download the packed_model.zip, I encountered an error:
batch response: This repository is over its data quota. Account responsible for LFS bandwidth should purchase more data packs to restore access.
Can you upload the ckpts to Google Drive to help us access it? Thanks a lot.
As an alternative, can you consider uploading the ckpts by releasing them? https://github.com/SamsungLabs/StepFormer/releases
Hi,
Thank you for your interest in our work. I am pursuing a deadline at the moment, will fix the saved model release asap. Hopefully by next week.
Regards,
Ran
On Thu, Sep 21, 2023 at 11:58 AM Shuhuai Ren @.***> wrote:
Updated the saved model in the repo. https://github.com/SamsungLabs/StepFormer/blob/main/packed_model.zip
Hi, when I use git lfs pull to download the packed_model.zip, I encountered an error:
batch response: This repository is over its data quota. Account responsible for LFS bandwidth should purchase more data packs to restore access.
Can you upload the ckpts to Google Drive to help us access it? Thanks a lot.
— Reply to this email directly, view it on GitHub https://github.com/SamsungLabs/StepFormer/issues/1#issuecomment-1729867022, or unsubscribe https://github.com/notifications/unsubscribe-auth/AHV5WILKRT2MLS7OW7B7KMDX3RP2FANCNFSM6AAAAAA4TL6ZYE . You are receiving this because you commented.Message ID: @.***>
Hi Ran, do you have a chance to fix the saved model release so that we can download it. Thanks!
Hi, our administrator tells me that the lfs quota should refresh periodically. Could you try again?
I tried git clone (also git lfs clone) the repo and the following error pops up:
batch response: This repository is over its data quota. Account responsible for LFS bandwidth should purchase more data packs to restore access. error: failed to fetch some objects from 'https://github.com/SamsungLabs/StepFormer.git/info/lfs'
@lavenderrz Can you just upload the ckpt to huggingface? https://huggingface.co/new. This is very convenient to operate, and it will not cause any cost.
I am trying to obtain permission for hosting the model somewhere else. I will send an update when I hear back. I apologize for limited capacity we have in this account.
Thanks @lavenderrz . Could you also share the train/test split on COIN and CrossTask so that we can try to compare with the reported results. As described in the supplementary material:
For direct comparison with StepFormer, we train all the baselines using the same video features [22], and fix the same training and testing splits in every dataset.
Hi @lavenderrz , could you please kindly share the train/test split on COIN/CrossTask. Are they simply following the standard train/test splits of the two datasets, or is there a customized one?
Hi @lavenderrz , could you please kindly share the train/test split on COIN/CrossTask. Are they simply following the standard train/test splits of the two datasets, or is there a customized one?
@xyang35 we follow the standard split for all dataset we tested on. Hope this helps.
Hello. Has anyone found a way to download the checkpoints? @lavenderrz @hadjisma
