How to create a custom Cross Validation method for TS

Hi guys, how to create a custom cv method compatible with GridSearchCV?
I would like to create a TSCV based on lag. I mean, a specific time-window to train/test adding/removing exogenous features. To illustrate, suppose I have the following data:

X = [[ 74.9429, 72.4737, 74.8085, 75.9004, 76.2531],
[ 72.4737, 74.8085, 75.9004, 76.2531, 73.7503],
[ 74.8085, 75.9004, 76.2531, 73.7503, 74.0778]]
y = [73.7503, 74.0778, 74.1702]

I create an array-like to use as cv in GridSearch:

tscv = [(array([ 74.9429, 72.4737, 74.8085, 75.9004, 76.2531, 139.2689]), 73.7503),
(array([ 72.4737, 74.8085, 75.9004, 76.2531, 73.7503, 140.3669]), 74.0778),
(array([ 74.8085, 75.9004, 76.2531, 73.7503, 74.0778, 139.6127]), 74.1702)]

Unfortunately it doesn’t work. I got the following error: “ValueError: No valid specification of the columns. Only a scalar, list or slice of all integers or all strings, or boolean mask is allowed”.

Am I doing something wrong?
Regards,