Error TrainUsing AutoML with Use Location Embeddings checked

[d3netxer]

ERROR 260203: An error occurred while preparing the data to train AutoML.

My ArcGIS Pro and Deep Learning Libraries are version 3.6

It runs fine when Use Location Embeddings is unchecked

I tested with a point dataset projected to UTM and also with WGS84

[scw]

@d3netxer that error indicates there was an exception raised during the preparation of the data, but the error message itself doesn't provide the underlying exception to evaluate what the problem is, the exception is currently written as:

        except Exception as e:
            arcpy.AddIDMessage("ERROR", 260203)
            exit()

If you're familiar with Python, you could try changing that in the tool.script.execute.py method to see the true exception.

@priyankatuteja can you evaluate this issue and how you'd like to have it be tracked? It isn't a framework issue so should likely live elsewhere.

[priyankatuteja]

@d3netxer Could you share the steps to reproduce the issue?

[d3netxer]

Hello, At first I couldn't change anything with the Python file because I don't have permissions to files on my C drive. However I was able to re-install ArcGIS Pro on my D drive permissions.

Anyhow, I was able to determine that the tool runs if Use Location Embeddings is unchecked, but it does not run using all the exact same settings if Location Embeddings is checked. Any idea why this is the case?

[pr-dev-del]

Hello @d3netxer , could you please provide your data and code/screenshots so that we could reproduce this on our end? We did try using location embeddings with a few datasets and the workflow seems to work fine.

[d3netxer]

@pr-dev-del is it possible to e-mail you a sanitized version of the dataset I am using?

[pr-dev-del]

@d3netxer yes, sure.

[tgkap]

ok thanks, I have just e-mailed you the dataset @pr-dev-del