Reproducibility

[wistuba]

Hey, thanks for sharing your code. I am currently trying to reproduce the numbers reported in the paper for S-iPrompts, but I'm getting very different numbers (sometimes worse, sometimes much better). Would you mind sharing the missing configs?

Thank you!

[4vicii]

hi，did you encounter a bug about models/sinet.py line 46:

pretrained_cfg = resolve_pretrained_cfg(variant, kwargs=kwargs)

pretrained_cfg = resolve_pretrained_cfg(variant, kwargs=kwargs) TypeError: resolve_pretrained_cfg() got an unexpected keyword argument 'kwargs'

[wistuba]

this repo doesn't seem to maintained nor in a shape that would allow for reproducing results. your error sounds like an issue with timm. these dependencies worked for me. give it a shot: pick your torch dependency with torch<1.13.2

numpy==1.21.2
pillow==9.0.1
timm==0.6.5
scipy==1.8.0
tqdm==4.63.0
scikit-learn==1.0.2
ftfy==6.1.1
regex==2022.3.15

refer also to #4 for some guidance on the right environment

[4vicii]

thks for your help.may you have a wonderful day 🥹🫡 saluteSent from my iPhoneOn Mar 17, 2024, at 20:32, wistuba @.***> wrote: this repo doesn't seem to maintained nor in a shape that would allow for reproducing results. your error sounds like an issue with timm. these dependencies worked for me. give it a shot: pick your torch dependency with torch<1.13.2 numpy==1.21.2 pillow==9.0.1 timm==0.6.5 scipy==1.8.0 tqdm==4.63.0 scikit-learn==1.0.2 ftfy==6.1.1 regex==2022.3.15

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

[4vicii]

Did you change the net_type slip to sip in cddb_sip.yaml?I match the dependencies to yours , but still get the bug 😔😔😔On Mar 17, 2024, at 22:06, PKU @.> wrote:thks for your help.may you have a wonderful day 🥹🫡 saluteSent from my iPhoneOn Mar 17, 2024, at 20:32, wistuba @.> wrote: this repo doesn't seem to maintained nor in a shape that would allow for reproducing results. your error sounds like an issue with timm. these dependencies worked for me. give it a shot: pick your torch dependency with torch<1.13.2 numpy==1.21.2 pillow==9.0.1 timm==0.6.5 scipy==1.8.0 tqdm==4.63.0 scikit-learn==1.0.2 ftfy==6.1.1 regex==2022.3.15

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