From 5343aef1c8300f2caba97928b9b8e16e8feb5f9c Mon Sep 17 00:00:00 2001
From: Xingjian Shi <xshiab@connect.ust.hk>
Date: Thu, 21 May 2020 09:06:47 -0700
Subject: [PATCH] [Numpy] Fix gluon activations (#18370)

* fix activation numpy bug

* Update test_numpy_gluon.py

* fix
---
 python/mxnet/gluon/nn/activations.py      | 18 +++++---
 tests/python/unittest/test_numpy_gluon.py | 50 +++++++++++++++++++++++
 2 files changed, 63 insertions(+), 5 deletions(-)

diff --git a/python/mxnet/gluon/nn/activations.py b/python/mxnet/gluon/nn/activations.py
index 1b9ce91dd2aa..3cccc851e39b 100644
--- a/python/mxnet/gluon/nn/activations.py
+++ b/python/mxnet/gluon/nn/activations.py
@@ -139,7 +139,8 @@ def __init__(self, alpha_initializer=initializer.Constant(0.25),
                                          init=alpha_initializer)
 
     def hybrid_forward(self, F, x, alpha):
-        return F.LeakyReLU(x, gamma=alpha, act_type='prelu', name='fwd')
+        leaky_relu = F.npx.leaky_relu if is_np_array() else F.LeakyReLU
+        return leaky_relu(x, gamma=alpha, act_type='prelu', name='fwd')
 
 
 class ELU(HybridBlock):
@@ -167,7 +168,8 @@ def __init__(self, alpha=1.0, **kwargs):
         self._alpha = alpha
 
     def hybrid_forward(self, F, x):
-        return F.LeakyReLU(x, act_type='elu', slope=self._alpha)
+        leaky_relu = F.npx.leaky_relu if is_np_array() else F.LeakyReLU
+        return leaky_relu(x, act_type='elu', slope=self._alpha)
 
 
 class SELU(HybridBlock):
@@ -187,7 +189,9 @@ def __init__(self, **kwargs):
         super(SELU, self).__init__(**kwargs)
 
     def hybrid_forward(self, F, x):
-        return F.LeakyReLU(x, act_type='selu', name='fwd')
+        leaky_relu = F.npx.leaky_relu if is_np_array() else F.LeakyReLU
+        return leaky_relu(x, act_type='selu', name='fwd')
+
 
 class GELU(HybridBlock):
     r"""
@@ -206,7 +210,8 @@ def __init__(self, **kwargs):
         super(GELU, self).__init__(**kwargs)
 
     def hybrid_forward(self, F, x):
-        return F.LeakyReLU(x, act_type='gelu', name='fwd')
+        leaky_relu = F.npx.leaky_relu if is_np_array() else F.LeakyReLU
+        return leaky_relu(x, act_type='gelu', name='fwd')
 
 
 class Swish(HybridBlock):
@@ -232,4 +237,7 @@ def __init__(self, beta=1.0, **kwargs):
         self._beta = beta
 
     def hybrid_forward(self, F, x):
-        return x * F.sigmoid(self._beta * x, name='fwd')
+        if is_np_array():
+            return x * F.npx.sigmoid(self._beta * x)
+        else:
+            return x * F.sigmoid(self._beta * x, name='fwd')
diff --git a/tests/python/unittest/test_numpy_gluon.py b/tests/python/unittest/test_numpy_gluon.py
index 713591795a43..d6fe236fe6f7 100644
--- a/tests/python/unittest/test_numpy_gluon.py
+++ b/tests/python/unittest/test_numpy_gluon.py
@@ -25,6 +25,7 @@
 import mxnet as mx
 from mxnet import gluon, autograd, np
 from mxnet.test_utils import use_np, assert_almost_equal, check_gluon_hybridize_consistency
+from mxnet.gluon import nn
 from common import with_seed
 import random
 
@@ -431,3 +432,52 @@ def hybrid_forward(self, F, valid_length):
 
     assert mx.test_utils.same(out1.asnumpy(), out2.asnumpy())
 
+
+@with_seed()
+@use_np
+def test_activations_leakyrelu():
+    # Currently, all the activation tests, we will just test for runnable.
+    act_layer = nn.LeakyReLU(0.1)
+    out = act_layer(mx.np.random.uniform(size=(10,)))
+    out.asnumpy()
+
+
+@with_seed()
+@use_np
+def test_activations_prelu():
+    act_layer = nn.PReLU()
+    act_layer.initialize()
+    out = act_layer(mx.np.random.uniform(size=(10,)))
+    out.asnumpy()
+
+
+@with_seed()
+@use_np
+def test_activations_elu():
+    act_layer = nn.ELU(1.0)
+    out = act_layer(mx.np.random.uniform(size=(10,)))
+    out.asnumpy()
+
+
+@with_seed()
+@use_np
+def test_activations_selu():
+    act_layer = nn.SELU()
+    out = act_layer(mx.np.random.uniform(size=(10,)))
+    out.asnumpy()
+
+
+@with_seed()
+@use_np
+def test_activations_gelu():
+    act_layer = nn.GELU()
+    out = act_layer(mx.np.random.uniform(size=(10,)))
+    out.asnumpy()
+
+
+@with_seed()
+@use_np
+def test_activations_swish():
+    act_layer = nn.Swish()
+    out = act_layer(mx.np.random.uniform(size=(10,)))
+    out.asnumpy()