Azure Artifact Referencing

Following doesn’t work for us. Is their any way to solve the following problems for Artifact reference with Azure Blob Storage:

  1. How should we pass credentials to wandb ? As for Amazon S3 and GCS the priority and env variables are mentioned in docs.

  2. It was recommended that by passing az:// as prefix will work similar to whats done s3 bucket and gcs. However I didn’t see any storage handler in wandb code for azure. I wonder how would it work just by passing a prefix ? Furthermore, unlike boto for s3 and google-cloud-storage sdk for gcs. I don’t see any requirement of azure-storage in requirements.txt. Is their any Microsoft Azure Storage SDK for Python somewhere in code that I can not find ??

  3. Although it doesn’t make any sense still I gave it a try, and as expected. Following are the results.

run = wandb.init(project="Dummy_Training", job_type="upload")
bucket = 'az://azurestorage.blob.core.windows.net/container_name'
dataset_at = wandb.Artifact('sample',type="raw_data")

dataset_at.add_reference(bucket)
run.log_artifact(dataset_at)
run.finish()

And I get the following Error

  1. Do I need to pass something in name parameter? What would be the entry name for azure?
    Seems like az:// is defiantly not in your known handlers

Is their any way for Azure Artifact Referencing (azure blob storage) to work. And please let me know if their is any thing that I am missing. Any example for the resolution of this problem will be much appreciated.

Thanks.

And I also didn’t see any storage handler in wandb code for azure.
C1BCD638F8B84BF0A7F85024A39F405A

And I also didnt see a way to pass azure credential to wandb in the artifact referencing document.
F9B220A31E5648CCA496FEE206CDCF27

Hi @hassanw65. Unfortunately Azure reference artifacts aren’t supported currently. We have a ticket created with our engineering team and increased the priority for it. I will let you know when there are updates.