diff --git a/pymc/logprob/abstract.py b/pymc/logprob/abstract.py
index 5c7f28e661..62bf4999a6 100644
--- a/pymc/logprob/abstract.py
+++ b/pymc/logprob/abstract.py
@@ -242,7 +242,8 @@ def make_node(self, rv, value):
         return Apply(self, [rv, value], [rv.type(name=rv.name)])
 
     def perform(self, node, inputs, out):
-        raise NotImplementedError("ValuedVar should not be present in the final graph!")
+        warnings.warn("ValuedVar should not be present in the final graph!")
+        out[0][0] = inputs[0]
 
     def infer_shape(self, fgraph, node, input_shapes):
         return [input_shapes[0]]
diff --git a/pymc/logprob/transform_value.py b/pymc/logprob/transform_value.py
index 4a28d5cd4a..2e1b96d343 100644
--- a/pymc/logprob/transform_value.py
+++ b/pymc/logprob/transform_value.py
@@ -11,6 +11,7 @@
 #   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 #   See the License for the specific language governing permissions and
 #   limitations under the License.
+import warnings
 
 from collections.abc import Sequence
 
@@ -40,7 +41,8 @@ def make_node(self, tran_value: TensorVariable, value: TensorVariable):
         return Apply(self, [tran_value, value], [tran_value.type()])
 
     def perform(self, node, inputs, outputs):
-        raise NotImplementedError("These `Op`s should be removed from graphs used for computation.")
+        warnings.warn("TransformedValue should not be present in the final graph!")
+        outputs[0][0] = inputs[0]
 
     def infer_shape(self, fgraph, node, input_shapes):
         return [input_shapes[0]]
diff --git a/tests/logprob/test_basic.py b/tests/logprob/test_basic.py
index 64cbf63b3e..3117d8f1aa 100644
--- a/tests/logprob/test_basic.py
+++ b/tests/logprob/test_basic.py
@@ -436,3 +436,26 @@ def test_ir_rewrite_does_not_disconnect_valued_rvs():
         logp_b.eval({a_value: np.pi, b_value: np.e}),
         stats.norm.logpdf(np.e, np.pi * 8, 1),
     )
+
+
+def test_ir_ops_can_be_evaluated_with_warning():
+    _eval_values = [None, None]
+
+    def my_logp(value, lam):
+        nonlocal _eval_values
+        _eval_values[0] = value.eval()
+        _eval_values[1] = lam.eval({"lam_log__": -1.5})
+        return value * lam
+
+    with pm.Model() as m:
+        lam = pm.Exponential("lam")
+        pm.CustomDist("y", lam, logp=my_logp, observed=[0, 1, 2])
+
+    with pytest.warns(
+        UserWarning, match="TransformedValue should not be present in the final graph"
+    ):
+        with pytest.warns(UserWarning, match="ValuedVar should not be present in the final graph"):
+            m.logp()
+
+    assert _eval_values[0].sum() == 3
+    assert _eval_values[1] == np.exp(-1.5)