search

xuebinqin / U-2-Net

The code for our newly accepted paper in Pattern Recognition 2020: "U^2-Net: Going Deeper with Nested U-Structure for Salient Object Detection."
Apache License 2.0
8.34k stars 1.43k forks source link

Training with more than one GPU... #79

Open h-peterseim opened 3 years ago

h-peterseim commented 3 years ago

...would be a nice feature. Is it possible to deliver the needed scripts? Train on an 1080TI with 14 images batch size, but the need in time is huge if there are many images in the training set.

xuebinqin commented 3 years ago

Thanks for your suggestion. But currently we don't have that many GPUs ? If you have, you can give it a try and put a pull request.

On Sat, Oct 10, 2020 at 4:51 AM h-peterseim notifications@github.com wrote:

...would be a nice feature. Is it possible to deliver the needed scripts? Train on an 1080TI with 14 images batch size, but the need in time is huge if there are many images in the training set.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/NathanUA/U-2-Net/issues/79, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADSGORPNM3OOEXJPS2PFUZLSKA4DDANCNFSM4SK7RICQ .

-- Xuebin Qin PhD Department of Computing Science University of Alberta, Edmonton, AB, Canada Homepage:https://webdocs.cs.ualberta.ca/~xuebin/

h-peterseim commented 3 years ago

Uhh, i'm an python noob, no idea what's to do to utilize more than one GPU. I'm currently renting machine via an cloud service, you can test there for free because ob 50$ free credits @ signup. So the machine is not an problem, and if needed i can provide an 2 GPU machine for testing reasons too. I've wrote an manual what's to do to get an usable machine there (manual is in german) here : https://www.petit-souris.de/?beitrag=35

flower-Ding commented 1 year ago

Thanks for your suggestion. But currently we don't have that many GPUs ? If you have, you can give it a try and put a pull request. … On Sat, Oct 10, 2020 at 4:51 AM h-peterseim @.***> wrote: ...would be a nice feature. Is it possible to deliver the needed scripts? Train on an 1080TI with 14 images batch size, but the need in time is huge if there are many images in the training set. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#79>, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADSGORPNM3OOEXJPS2PFUZLSKA4DDANCNFSM4SK7RICQ . -- Xuebin Qin PhD Department of Computing Science University of Alberta, Edmonton, AB, Canada Homepage:https://webdocs.cs.ualberta.ca/~xuebin/

have you found some methods to train with multi-gpus?