add additional logging to test failures to help catch flakey inducing…

… point tests (#616)

Summary:
Pull Request resolved: #616

Can't replicate test failure even with the buck command, so just add some extra info to help catch failures in the future.

Reviewed By: crasanders

Differential Revision: D69138310

fbshipit-source-id: 916c3ece02169c1866318328cdd41cc3db392a0c

tests/test_points_allocators.py

@@ -101,7 +101,9 @@ def test_kmeans_allocator_allocate_inducing_points(self):
 
         self.assertIs(model.inducing_point_method.last_allocator_used, KMeansAllocator)
         inducing_points = model.variational_strategy.inducing_points
-        self.assertTrue(inducing_points.shape == (9, 2))
+        self.assertTrue(
+            inducing_points.shape == (9, 2), f"shape is {inducing_points.shape}"
+        )
         # We made ints, so mod 1 should be 0s, so we know these were the original inputs
         self.assertTrue(torch.all(inducing_points % 1 == 0))