Predicting single floats based on set of 2 feature arrays each of 100 values

I am trying to predict audio to video desynchronization based on set of two arrays of lenght 100 which consist of coresponding audio and video samples.

The problem is that my labels are single floats (values of shift), while both audio and video data are arrays of lenght 100. So far I tried Lasso for that problem but I couldn't get rid of errors while fitting model.

This is how my data looks like:

 print(audio)
[[0.675324 ... 0.59183673, ] ... [0.34116661 ... 0.12759797]]

 print(audio.shape)
(200, 100)

 print(video)
[[0.67532086 ... 0.12184522 ][0.34116661 ... 0.10322892]]

 print(video.shape)
(200, 100)

 print(shift)
[ 900  500  ...   50 -250]

 print(shift.shape)
(200,)

Is there any way to do that with that model or maybe there is some other solution to that?