Kaggle Not Using GPU At ALL


OK, so i switch the runtime to use The GPU and restart the notebook.

I am using the below git-hub project to remove the background from images .

It takes approx 30 mins to remove background of 86 Images.

I have used Google drive to read my Images (It took the Same 30 mins)

Then i downloaded the same dateset to my Google colab/ Kaggle itself from MediaFire (Again took same 30 mins)

Now the thing is without the GPU, it also takes the Same time, that is 30 Mins (SO am i using the GPU or CPU wrong or is it not being utilized properly. ?

Please Help, Thanks

Proof that it only Uses FULL 100% CPU and Barely Any GPU (I am using u2net Model)

Which script in the linked repository are you using and are you moving the model as well as the data to the GPU manually?

Hi. So I edited the requirements file and removed all the == signs. So installed all the latest version.

And the following code I used in kaggle/ google Collab.

Please Help. Thanks :slight_smile:

It doesn’t seem you are installing PyTorch at all but MXNet so the MXNet discussion board might be a better place to ask. :wink:

hi in the github project above, it says that if using pytorch (u2net, u2netp), so i am using utnet. sorry a bit newbie if confused

can you try this out in google collab or kaggle if you have time

sorry and thanks :slight_smile: