Loop fusion failed

Sure, the effect is additive (it is an initial approach in order to figure out how to handle shaping and indexing). But what is the shape of meanRunner? I have 108 runner so the first dimension will be 108, but not all of them have a race mean for every year because they haven’t run in every year. The years are 7.
If i write something like
meanRunner = pm.StudentT(“meanRunner”, nu=3, mu=meanYear,sigma=240,shape=(108,7))
I don’t know how the ‘empty’ means interfere with the inference.
So i was looping over the runners and i was creating 108 different variables with proper size for each runner.
e.g.
meanRunner1 = pm.StudentT(“meanRunner1”, nu=3, mu=meanYear,sigma=240,shape=3)
if runner1 had races in 3 years
meanRunner2 = pm.StudentT(“meanRunner2”, nu=3, mu=meanYear,sigma=240,shape=5)
if runner2 had races in 5 years.
The sampling was very slow but the results were good. So i am asking how i can handle such a situation in an effective way. I don’t have experience in working with tensors.