Azure ML Studio can't generate job when datastore is used as input; failed string validation? Also, code files not being loaded on job.

Question

The error I get:
Job submission error
UserError: Unexpected JobInputType in request body: [UriFolder]. Supported types: literal,uri_file,uri_folder,mltable,custom_model,mlflow_model,triton_model.
Trace ID : d32c0fca-df85-4679-854d-f9b76989bb5cClient request ID : be63cb66-373b-457f-82d5-43d6dd132eb6Service request ID : 3c5afbfe-2994-48f8-8379-d2868eefb878

It seems to me that the form provides the string "UriFolder" to tell what type of data is being submitted, but it fails validation because it's checked against "uri_folder". My best guess is that the Azure ML Studio frontend is providing an invalid string for the backend, which fails validation.

More details about my issue:
I managed to load my code into blob storage. Now I'm trying to create a job, but whenever I add to the job any input or output which uses a Datastore (such as workspaceworkingdirectory where my data is, or a blob), the job fails to generate. I'm using Azure ML Studio (web interface).

Settings when creating the input for the job:

Input type = Data
data type = folder
data source = datastore
selected datastore = workspaceworkingdirectory
path to data = Users/me/data/** (chosen using the "browse" button, so it must be correct)
input can be mounted in any way (read only, download, direct), I always get the same error. Distributed training settings also don't matter, anything I choose gives the same error.

Edit: I found yet another error in Azure ML Studio.

To circumvent the error reported above, I've painstakingly uploaded all my data to the blob storage container, so I wouldn't have to use any inputs or outputs to create the job. My strategy worked, the job was created, but it fails at start with the error message "detector.py: file not found". I've added an "ls -R" to the beginning of the command that is executed to start the job, and then checked the logs. The "ls" shows that there are absolutely no files on the path where the terminal is started, only 4 empty folders: azureml-logs, logs, outputs, user_logs.
I've tried uploading the code from my PC through the form when creating the job, it takes longer to create the job because the files are being uploaded, but I still get the same error.

Where are my files going? Why I don't get an error code telling me there was a problem accessing/transfering the data? Or is the data actually being transfered, but the terminal is pointing to the wrong folder? The yaml file generated by the web GUI is suspiciously dry, with no mentions to code path or code files...

Are these two bugs in the Azure ML Studio known? Do you maybe have any open positions for a test engineer? ;-)
Both problems I've reported happen in the same page shown in the following screenshot.

Answer 1

Same problem. Job create through UI. Just to be sure I did it both as folder and file. Seems not to be an issue when doing it through Azure CLI (there I had a docker fail, but that is another issue)