search

mit-quest / necstlab-damage-segmentation

MIT License
5 stars 6 forks source link

No error messages raised when dataset name is not found #87

Closed BrendenBarbour closed 3 years ago

BrendenBarbour commented 3 years ago

Causes various errors when a dataset that does not exist is called in multiple locations, train_segmentation_model, etc.

CarolinaFurtado commented 3 years ago

what to take this one, @BrendenBarbour? I think doing something like: 

try:
    copy_folder_locally_if_missing(os.path.join(gcp_bucket, 'datasets', train_config['dataset_id']),
                                   local_dataset_dir)
except ...
     some type of error

in train_segmentation_models.py does the trick. Maybe we could take this opportunity to do the other steps of the workflow.

If you are busy with other things, I can tackle these.

CarolinaFurtado commented 3 years ago

This will be addressed in issue #91