[Bug]: IndexError: index 1 is out of bounds for axis 0 with size 1, raised at: /opt/conda/lib/python3.9/site-packages/mindsdb/integrations/libs/learn_process.py#183 about mindsdb HOT 3 CLOSED

@pedrofluxa
Can you have a look please ^^

from mindsdb.

your data unit is same in all rows...

from mindsdb.

Hi @niyascode007

The reason you are seeing this error is that every row in your table has unit=90, as @ZaKull correctly pointed out.

This makes type_infer suggest to dataprep_ml that this column should be dropped, which in turns causes the time-series pipeline to fail and throw that rather obscure error message.

I modified the file you uploaded to provide you with a working example. Note the following

• Timestamps are now separated by a second
• Timestamps don't have time-zone information (it triggers nasty warnings from pandas) As a personal advice, try to always save your timestamps in UTC.
• Column unit now variates. I would like you to try setting most values to 90 and just a few to different ones to see if it works. My gut feeling is that it won't.

Here is the SQL statements for you to try them out (after uploading the file and name it "data")

--- The following code creates a Generative AI table based on a Time-series model provided by Lightwood.
--- The model requires as input the last 4 measurements and predicts the next two.
---
--- Instructions
--- > Upload "sensor_data_fixed.csv" to MindsDB and name it "data" using the GUI.
--- > Copy and paste this script, execute.


SHOW TABLES FROM files;

--- These files can be queried as tables,
--- You just uploaded sensor_data_non_constant_unit

SELECT * FROM files.data LIMIT 10;

DROP MODEL mindsdb.model66;

CREATE MODEL mindsdb.model66
FROM files
(SELECT * from data) 
PREDICT unit
ORDER BY timestamp
WINDOW 4
HORIZON 2;

DESCRIBE mindsdb.model66;

SELECT m.timestamp as date, m.unit as forecast
  FROM mindsdb.model66 as m 
    JOIN files.data as t
    WHERE t.timestamp > '2023-12-10';

sensor_data_fixed.csv

Good luck!

from mindsdb.