Hello,
Yes, clearly I needed to change that,
I now have @theano.compile.ops.as_op(itypes=[t.dvector,t.dscalar,t.dscalar],otypes=[t.dscalar])
nevertheless I still get the exact same error (which is why I oversaw this simple mistake)…
I did some tests, and it seems that the keyword argument in the error is the first one that is passed in observed. If I change the order of the dictionnary, by putting alpha first, it yields
TypeError: make_node() got an unexpected keyword argument 'alpha'
Going into the code of op.py it seems that the kwargs are maybe confusing make_node() here.
I tried to remove the as_op of my proc_test function, then it runs without this error, but instead of using the value of alpha it sees it as a string in the function, which of course produces errors later on.
I’m not sure if I’m using this DensityDist the right way…