There are several convolution functions in pytensor.tensor.conv, but this module is not very well documented nor actively maintained. You will need to do some investigation and see if any of them suit your purpose. If none do, you can use a scan to implement exactly the behavior you desire.
We’re very eager for documentation contributions, though. If you do some research and want to contribute some explanations on how each function works in the form of docstrings or examples, please don’t hesitate to open a PR.
On that note, there’s a PyData Sprint for first-time contributions, so you can come meet the community team over the next 2 days and get some pointers on how to contribute. See here and here for our entries on the schedule.