Fix dtype casting bug in icdf function

Author: ricardoV94

pymc/logprob/basic.py

@@ -127,7 +127,7 @@ def icdf(
     rv: TensorVariable, value: TensorLike, warn_missing_rvs: bool = True, **kwargs
 ) -> TensorVariable:
     """Create a graph for the inverse CDF of a  Random Variable."""
-    value = pt.as_tensor_variable(value, dtype=rv.dtype)
+    value = pt.as_tensor_variable(value, dtype="floatX")
     try:
         return _icdf_helper(rv, value, **kwargs)
     except NotImplementedError:

tests/logprob/test_basic.py

@@ -510,3 +510,14 @@ def test_warn_random_found_probability_inference(func, scipy_func, test_value):
                 test_value
             ),
         )
+
+
+def test_icdf_discrete():
+    p = 0.1
+    value = 0.9
+    dist = pm.Geometric.dist(p=p)
+    dist_icdf = icdf(dist, value)
+    np.testing.assert_almost_equal(
+        dist_icdf.eval(),
+        sp.geom.ppf(value, p),
+    )