[GitHub] eric-haibin-lin commented on a change in pull request #10913: Fix _backward_norm op registration

[GitBox <gi...@apache.org>]

eric-haibin-lin commented on a change in pull request #10913: Fix _backward_norm op registration
URL: https://github.com/apache/incubator-mxnet/pull/10913#discussion_r187806622
 
 

 ##########
 File path: tests/python/unittest/test_sparse_ndarray.py
 ##########
 @@ -918,18 +918,22 @@ def test_sparse_nd_check_format():
 
 @with_seed()
 def test_sparse_nd_norm():
-    def check_sparse_nd_norm(stype, shape, density):
+    def check_sparse_nd_norm(stype, shape, density, **kwargs):
         data, _ = rand_sparse_ndarray(shape, stype, density)
-        norm = data.norm()
-        expected_norm = np.linalg.norm(data.asnumpy())
-        assert_almost_equal(norm.asnumpy(), expected_norm)
+        norm = data.norm(**kwargs)
+        expected_norm = data.tostype('default').norm(**kwargs)
+        assert_almost_equal(norm.asnumpy(), expected_norm.asnumpy())
 
     shape = (5, 5)
     stypes = ['row_sparse', 'csr']
-    densities = [0, 0.5]
+    densities = [0, 0.5, 1]
     for stype in stypes:
         for density in densities:
-            check_sparse_nd_norm(stype, shape, density)
+           check_sparse_nd_norm(stype, shape, density, axis=None, keepdims=False, ord=2)
+
+    # test fallback
+    check_sparse_nd_norm(stype, shape, density, axis=0, keepdims=False, ord=2)
+    check_sparse_nd_norm(stype, shape, density, axis=None, keepdims=True, ord=2)
 
 Review comment:
   i had this before but even for dense ord=2 is not implemented 

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services