AI Fabric - MLSkill Test inside Studio - Error

Help AI Center
1

I’m already deploy my own .zip with the ML Package. Inside the .zip i have all required files (main.py, etc…). On AI Fabric the deploy ML Skill was OK. Now I can see inside Studio my ML_Skill. No far so good.

After that I put MLSkill Activity inside the my sequence and select my ML_Skill. When i click in Test Skill button, and enter the input for my ML, i’m always got an ERROR.

My ML has JSON input format.

Example = ‘{“COLUMN_STRING1”: “ABC”, “COLUMN_STRING2”: “XYZ”, “COLUMN_INT1”: 1, “COLUMN_INT2”: 2}’

This format is correct? What’s the correct way?

2

If i run a test inside main.py using the format
‘{“COLUMN_STRING1”: “ABC”, “COLUMN_STRING2”: “XYZ”, “COLUMN_INT1”: 1, “COLUMN_INT2”: 2}’
The Python Console returns correct!

3

Hi @ti1
If you are using the test Skill pannel, the input is already String so no need for simple quote arround. This should work:
{"COLUMN_STRING1": "ABC", "COLUMN_STRING2": "XYZ", "COLUMN_INT1": 1, "COLUMN_INT2": 2}
On the activity directly you need to send a string and escape the double quotes:
"{""COLUMN_STRING1"": ""ABC"", ""COLUMN_STRING2"": ""XYZ"", ""COLUMN_INT1"": 1, ""COLUMN_INT2"": 2}"

Jeremy

4

Thanks for your help Jeremy
But when i try this, i got an error:

if ((self._feature_names_in is None or feature_names is None)\nAttributeError: ‘ColumnTransformer’ object has no attribute ‘_feature_names_in’",
“trace_id”: null

regards

5

Jeremy. The error above was inside the “Test Skill”. If a run to role sequence, passing a string to ML Skill Activity, the error is different

Extra Data: line 1 column 4 (char3)
“trace_id”: nullError_MLSkill

6

Could you send us exact input that you are sending in both cases? Is there any chance you also share your ML Package with us (you can remove the model itself if needed). It would help us understand what’s going on, first error seems to indicate that you send right data but something is wrong in code, second seems to indicate that the String is not well formated.
Feel free to send me direct message if you prefer.

7

Jeremy,

Just to record here the solution, as you request, i review the requirements.txt and made some changes, pointing the exactly version of all python packages. Before this review i had some packages using older version and using “=>”

After that i can use the ML Skill with success!

Thanks a lot @Jeremy_Tederry it was really great!!

1 Like
8

This topic was automatically closed 3 days after the last reply. New replies are no longer allowed.