[incubator-mxnet] branch szha-patch-1 updated (1ef0a43 -> 0bfdd39)

[zh...@apache.org]

This is an automated email from the ASF dual-hosted git repository.

zhasheng pushed a change to branch szha-patch-1
in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git.


    omit 1ef0a43  Fix nadam
     new 0bfdd39  Fix nadam

This update added new revisions after undoing existing revisions.
That is to say, some revisions that were in the old version of the
branch are not in the new version.  This situation occurs
when a user --force pushes a change and generates a repository
containing something like this:

 * -- * -- B -- O -- O -- O   (1ef0a43)
            \
             N -- N -- N   refs/heads/szha-patch-1 (0bfdd39)

You should already have received notification emails for all of the O
revisions, and so the following emails describe only the N revisions
from the common base, B.

Any revisions marked "omit" are not gone; other references still
refer to them.  Any revisions marked "discard" are gone forever.

The 1 revisions listed above as "new" are entirely new to this
repository and will be described in separate emails.  The revisions
listed as "add" were already present in the repository and have only
been added to this reference.


Summary of changes:
 tests/python/unittest/test_optimizer.py | 28 ++++++++++++++++++++++++++++
 1 file changed, 28 insertions(+)

-- 
To stop receiving notification emails like this one, please contact
['"commits@mxnet.apache.org" <co...@mxnet.apache.org>'].

[incubator-mxnet] 01/01: Fix nadam

[zh...@apache.org]

This is an automated email from the ASF dual-hosted git repository.

zhasheng pushed a commit to branch szha-patch-1
in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git

commit 0bfdd39bb862bb94b13370e88c402ce52fb5d1b3
Author: Sheng Zha <sz...@users.noreply.github.com>
AuthorDate: Mon Dec 18 13:08:18 2017 -0800

    Fix nadam
---
 python/mxnet/optimizer.py               |  2 +-
 tests/python/unittest/test_optimizer.py | 28 ++++++++++++++++++++++++++++
 2 files changed, 29 insertions(+), 1 deletion(-)

diff --git a/python/mxnet/optimizer.py b/python/mxnet/optimizer.py
index 2da6452..7e8e7c2 100644
--- a/python/mxnet/optimizer.py
+++ b/python/mxnet/optimizer.py
@@ -1098,7 +1098,7 @@ class Nadam(Optimizer):
         t = self._index_update_count[index]
 
         # preprocess grad
-        grad *= self.rescale_grad + wd * weight
+        grad = grad * self.rescale_grad + wd * weight
         if self.clip_gradient is not None:
             grad = clip(grad, -self.clip_gradient, self.clip_gradient)
 
diff --git a/tests/python/unittest/test_optimizer.py b/tests/python/unittest/test_optimizer.py
index ec4fbfd..921c1a9 100644
--- a/tests/python/unittest/test_optimizer.py
+++ b/tests/python/unittest/test_optimizer.py
@@ -18,6 +18,7 @@
 import numpy as np
 import mxnet as mx
 import mxnet.lr_scheduler as lr_scheduler
+from mxnet import gluon
 import unittest
 from nose.tools import raises
 import math
@@ -644,6 +645,33 @@ def test_ftrl():
         compare_optimizer(opt1(sparse_update=True, **kwarg), opt2(**kwarg), shape,
                           np.float32, w_stype='row_sparse', g_stype='row_sparse')
 
+def test_nadam():
+
+    def get_net(num_hidden, flatten=True):
+        data = mx.symbol.Variable('data')
+        fc1 = mx.symbol.FullyConnected(data, name='fc1', num_hidden=128, flatten=flatten)
+        act1 = mx.symbol.Activation(fc1, name='relu1', act_type="relu")
+        fc2 = mx.symbol.FullyConnected(act1, name = 'fc2', num_hidden = 64, flatten=flatten)
+        act2 = mx.symbol.Activation(fc2, name='relu2', act_type="relu")
+        fc3 = mx.symbol.FullyConnected(act2, name='fc3', num_hidden=num_hidden, flatten=flatten)
+        return fc3
+    np.random.seed(1234)
+    N = 20
+    data = mx.random.uniform(-1, 1, shape=(N, 10))
+    label = mx.random.uniform(-1, 1, shape=(N, 1))
+    data_iter = mx.io.NDArrayIter(data, label, batch_size=10, label_name='label', shuffle=True)
+    output = get_net(1)
+    l = mx.symbol.Variable('label')
+    Loss = gluon.loss.L1Loss()
+    loss = Loss(output, l)
+    loss = mx.sym.make_loss(loss)
+    mod = mx.mod.Module(loss, data_names=('data',), label_names=('label',))
+    mod.fit(data_iter, num_epoch=200, optimizer_params={'learning_rate': 0.01},
+            initializer=mx.init.Xavier(magnitude=2), eval_metric=mx.metric.Loss(),
+            optimizer='nadam')
+    assert mod.score(data_iter, eval_metric=mx.metric.Loss())[0][1] < 0.1
+
+
 if __name__ == '__main__':
     import nose
     nose.runmodule()

-- 
To stop receiving notification emails like this one, please contact
"commits@mxnet.apache.org" <co...@mxnet.apache.org>.