[AIR] computer_vision_test is too big

[gjoliver]

Description

our ray_air/doc_code/computer_vision.py test is too heavy.
it runs 5 batch inference jobs in both tf and torch in a loop, and is taking longer than 15mins to finish.
we changed it to "enormous", to unblock the release for now.

it would be nice to reduce the size while still testing the functionalities.
for example, use a much smaller dataet, or run this on GPU node, etc.
the problem with large tests is that our CI takes a long time to finish.

cc @richardliaw

Link

https://github.com/ray-project/ray/blob/master/doc/source/ray-air/doc_code/computer_vision.py

[bveeramani]

Thanks! The dataset size is limited to 32 rows, so I'm not how much using a smaller dataset will help, but I'll take a look.

[richardliaw]

how about let’s run it on the GPU runners?

…

On Tue, May 23, 2023 at 5:14 PM Balaji Veeramani ***@***.***> wrote: Thanks! The dataset size is limited to 32 rows, so I'm not how much using a smaller dataset will help, but I'll take a look. — Reply to this email directly, view it on GitHub <#35692 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABCRZZI2PJPFP4MWGQS4XGTXHVHHVANCNFSM6AAAAAAYMSKXJU> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[bveeramani]

Yeah, no harm in doing that.