fastdup.run() takes a long time to complete

[dnevo]

It would be nice if you can add a progress bar.

[visualdatabase]

hi @dnevo there is a progress bar if you are using regular python and not a jupyter notebook. can you try it and let us know if this works for you?

[dnevo]

I am using mostly Google Colab - it may be useful if you can add a progress bar for it..

[dnth]

Hello @dnevo are you still facing the same issue?

[dnevo]

I haven't used it recently...

…

On Wed, Apr 19, 2023 at 4:45 AM Dickson Neoh ***@***.***> wrote: Hello @dnevo <https://github.com/dnevo> are you still facing the same issue? — Reply to this email directly, view it on GitHub <#23 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFOIAD55GDTC56F4EWUTOBLXB47VNANCNFSM6AAAAAAR3SWROQ> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[woctezuma]

I have run the notebook on Google Colab after reading about it in an issue about DINOv2:

• Export to ONNX supported? facebookresearch/dinov2#19 (comment)

Apparently, it tooks 30 minutes, which is much more than advertised for .run() with the default arguments. 😅

Is it because of the use of DINOv2 instead of the default arguments?
Is there a possibility to use this Python package with a GPU if a free GPU is available on Colab?

[dbickson]

HI @woctezuma please note that the mentioned runtime are for fatsdup model, which is lightweight model and not dino v2. This colab machine is very weak just 2 cores. We recommend 32 core machine to give you x16 times speedup so you can run on this dataset in around 2 minutes. We will consider releasing gpu version in the coming days.