You can do 1d convolution using conv2d, you just have to add a dummy dimension then remove it when you’re done. I don’t have any PyMC or PyTensor examples of this, but I found this discussion about 1d convolution from the Theano team, which linked to these examples in the Lasagna library. Everything should work exactly the same (this part of the code base hasn’t changed since back then afaik), you’ll just have to rename some imports.
The convolution library is a bit of an orphan from the days when Theano was all about neural networks. There used to be comprehensive docs for it but they seem to have gotten lost in the shuffle. I found them on the wayback machine, though, along with the API docs here. Maybe you could rescue it 