[GitHub] [incubator-mxnet] mikemwx commented on a change in pull request #15302: [Numpy] Numpy hstack

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

mikemwx commented on a change in pull request #15302: [Numpy] Numpy hstack
URL: https://github.com/apache/incubator-mxnet/pull/15302#discussion_r302873300
 
 

 ##########
 File path: tests/python/unittest/test_numpy_op.py
 ##########
 @@ -860,6 +860,72 @@ def get_new_shape(shape, axis):
                 assert_almost_equal(mx_out.asnumpy(), np_out, rtol=1e-3, atol=1e-5)
 
 
+@with_seed()
+@npx.use_np_shape
+def test_np_hstack():
+    class TestHStack(HybridBlock):
+        def __init__(self):
+            super(TestHStack, self).__init__()
+        
+        def hybrid_forward(self, F, a, *args):
+            return F.np.hstack([a] + list(args))
+
+    def get_new_shape(shape):
+        if len(shape) == 0:
+            l = random.randint(0,3)
+            if l == 0:
+                return shape 
+            else:
+                return (l,)
+        shape_lst = list(shape)
+        axis = 1 if len(shape) > 1 else 0
+        shape_lst[axis] = random.randint(0, 5)
+        return tuple(shape_lst)
+
+    shapes = [
+        (),
+        (1,),
+        (2,1),
+        (2,2,4),
+        (2,0,0),
+        (0,1,3),
+        (2,0,3),
+        (2,3,4,5)
+    ]
+    for hybridize in [True, False]:
+        for shape in shapes:
+            test_hstack = TestHStack()
+            if hybridize:
+                test_hstack.hybridize()
+            # test symbolic forward
+            a = np.random.uniform(size=get_new_shape(shape))
+            a.attach_grad()
+            b = np.random.uniform(size=get_new_shape(shape))
+            b.attach_grad()
+            c = np.random.uniform(size=get_new_shape(shape))
+            c.attach_grad()
+            d = np.random.uniform(size=get_new_shape(shape))
+            d.attach_grad()
+            with mx.autograd.record():
+                mx_out = test_hstack(a, b, c, d)
+            np_out = _np.hstack((a.asnumpy(), b.asnumpy(), c.asnumpy(), d.asnumpy()))
+            assert mx_out.shape == np_out.shape
+            assert_almost_equal(mx_out.asnumpy(), np_out, rtol=1e-3, atol=1e-5)
+
+            # test symbolic backward
+            mx_out.backward()
+            assert_almost_equal(a.grad.asnumpy(), _np.ones(a.shape), rtol=1e-3, atol=1e-5)
+            assert_almost_equal(b.grad.asnumpy(), _np.ones(b.shape), rtol=1e-3, atol=1e-5)
+            assert_almost_equal(c.grad.asnumpy(), _np.ones(c.shape), rtol=1e-3, atol=1e-5)
+            assert_almost_equal(d.grad.asnumpy(), _np.ones(d.shape), rtol=1e-3, atol=1e-5)
+
+            # test imperative
+            mx_out = np.hstack((a, b, c, d))
+            np_out = _np.hstack((a.asnumpy(),b.asnumpy(), c.asnumpy(), d.asnumpy()))
+            assert_almost_equal(mx_out.asnumpy(), np_out, rtol=1e-3, atol=1e-5)
+
+
 
 Review comment:
   Done

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to 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