future_regressor length not equal to training data length?!

[emobs]

In my training runs I keep getting this message in the debug log:
"future_regressor row count does not match length of training data"

However, both are definitely equal according to the logs.

"It is recommended that the .head(forecast_length) of both regressor_train and the df for training are dropped. df = df.iloc[forecast_length:]"

That's what I do and the resulting dataframes are equal in length.

What could be causing this?

Thanks for your reply in advance.

[emobs]

Some more information on this:

I got my 'series_data' dataframe, which is the data input for AutoTS. Works fine without a regressor.

My lagged regressor is created like this to include one single column (Confirmed) from the series_data:
`
regressor_data = series_data[['Confirmed']].copy()

regressor_train, regressor_forecast = create_lagged_regressor(
regressor_data,
forecast_length=ForecastLength,
frequency=f"{PredictionTimeframe}T",
scale=False,
summarize=None, #"auto",
backfill="ETS", #"bfill",
fill_na="pchip", #"spline",
)
regressor_train = regressor_train.iloc[ForecastLength:]
logger.debug(f"length regressor_train: {len(regressor_train)}")
series_data = series_data.iloc[ForecastLength:]
logger.debug(f"length series_data: {len(series_data)}")
`
The log shows the data lengths are equal.

When the model is run like this model.fit(series_data, future_regressor=regressor_train, weights=weights, result_file=result_file_path) then, the script fails with many (semlock) error messages.
When called without the regressor like this model.fit(series_data, weights=weights, result_file=result_file_path) the script completes successfully.

What could be the cause of this issue? Is the regressor not created correctly?