Tested exact Dirichlet.logp values againt scipy implementation

Given a mention in RELEASE-NOTES.md

RELEASE-NOTES.md

@@ -12,6 +12,7 @@
 - `Theano-PyMC v1.1.2` also fixed an important issue in `tt.switch` that affected the behavior of several PyMC distributions, including at least the `Bernoulli` and `TruncatedNormal` (see[#4448](https://github.com/pymc-devs/pymc3/pull/4448))
 - `math.log1mexp_numpy` no longer raises RuntimeWarning when given very small inputs. These were commonly observed during NUTS sampling (see [#4428](https://github.com/pymc-devs/pymc3/pull/4428)).
 - `ScalarSharedVariable` can now be used as an input to other RVs directly (see [#4445](https://github.com/pymc-devs/pymc3/pull/4445)).
+- Fixed `Dirichlet.logp` method to work with unit batch or event shapes (see [#4454](https://github.com/pymc-devs/pymc3/pull/4454)).
 
 ## PyMC3 3.11.0 (21 January 2021)
 

pymc3/tests/test_distributions.py

@@ -1694,8 +1694,10 @@ def test_dirichlet_with_batch_shapes(self, dist_shape):
         with pm.Model() as model:
             d = pm.Dirichlet("a", a=a)
 
-        value = d.tag.test_value
-        assert_almost_equal(dirichlet_logpdf(value, a), d.distribution.logp(value).eval().sum())
+        pymc3_res = d.distribution.logp(d.tag.test_value).eval()
+        for idx in np.ndindex(a.shape[:-1]):
+            scipy_res = scipy.stats.dirichlet(a[idx]).logpdf(d.tag.test_value[idx])
+            assert_almost_equal(pymc3_res[idx], scipy_res)
 
     def test_dirichlet_shape(self):
         a = tt.as_tensor_variable(np.r_[1, 2])