Google Cloud - handling data

Hi

i’ve been looking at some kaggle kennels and trying to replicate it on my gcloud jupyternotebook, but i have some problems with my images files and the training speed.

Images:
In kaggle there are uploaded as a zip and i dont do anything to unzip them, but i can access them anyways.
On gCloud i cant access the zip file and have to unzip it to make a databunch.
Is it in fact correct that i have to unzip it to use?

Speed:
I bought the p4 edition. But when i train i takes 2x time compared to when i run the same on kaggle which is a k80.
Anything i can do to get the same speed as kaggle or faster?

Hope somebody can help a newbie like me :smiley:

Cheers from Denmark