Skip to content

cuda version issue #14

Open
Open
@rajayarli

Description

@rajayarli

Hi, I am using Google Collab's v100gpu as my run time also, when I try to run the code I face the below error:
issue 1: version problem
image

the versions that you specified were not compatible with the Cuda version and torch py of the collab I tried to download the matching version but it still says bfloat16 does not support the Cuda. I even tried to use float16 instead of that still it is still persistent.

issue 2: dataset specific problem
There are a lot of troubles I faced like the dataset complexity which with only a few buildings with more than 250 GB of space.
image

I even tried to run them with a few buildings like this:
image
still, I am unable to run that always getting the index out of bounds, also I am unaware of the usage of semseg_classes. do we need to modify if we are using fewer buildings?

issue 3: will this approach work?
Right now I am trying to download the data set individually like rgb with all buildings and I will copy 10 images per building to save space. and download the rest of the buildings externally using wget since the omni dataset is ignoring some buildings I just want to ask for your help and guidance if possible.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions