Handling tensor with variable length of dimension

[kentarow-kawata]

Description

Hi PyTensor developpers,

I would like to work with pytensor to handle three dimensional arrays. The first and second dimensions are of fixed length, but the third dimension can have various lengths.

e.g.) 32n tensor
[ [ [0,1,2], [3,4], [5] ],
[ [6,7] , [8] , [9] ] ]

Is there any way to handle tensors with variable length of dimensions using pytensor? Or, is there any way to treat the example as a matrix with array data (numpy.array or list) as elements?