From bda06fa74bea9e5e4aef32019bd07b0178e53dc0 Mon Sep 17 00:00:00 2001 From: yuxiangw Date: Mon, 13 Aug 2018 14:18:12 -0700 Subject: [PATCH 1/5] documentation changes. added full reference --- python/mxnet/optimizer.py | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/python/mxnet/optimizer.py b/python/mxnet/optimizer.py index ab7dadb17a54..439b865aaee9 100644 --- a/python/mxnet/optimizer.py +++ b/python/mxnet/optimizer.py @@ -556,7 +556,11 @@ class Signum(Optimizer): state = momentum * state + (1-momentum)*rescaled_grad weight = (1 - lr * wd_lh) * weight - lr * sign(state) - See the original paper at: https://jeremybernste.in/projects/amazon/signum.pdf + Reference: + Jeremy Bernstein, Yu-Xiang Wang, Kamyar Azizzadenesheli & Anima Anandkumar. (2018). + signSGD: Compressed Optimisation for Non-Convex Problems. In ICML’18. + + See: https://arxiv.org/abs/1802.04434 For details of the update algorithm see :class:`~mxnet.ndarray.signsgd_update` and :class:`~mxnet.ndarray.signum_update`. From 39623b76d12ecd40ebee7103c25f33c6ae65b166 Mon Sep 17 00:00:00 2001 From: yuxiangw Date: Mon, 13 Aug 2018 14:27:59 -0700 Subject: [PATCH 2/5] fixing lint --- python/mxnet/optimizer.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/python/mxnet/optimizer.py b/python/mxnet/optimizer.py index 439b865aaee9..7be767d52174 100644 --- a/python/mxnet/optimizer.py +++ b/python/mxnet/optimizer.py @@ -556,8 +556,8 @@ class Signum(Optimizer): state = momentum * state + (1-momentum)*rescaled_grad weight = (1 - lr * wd_lh) * weight - lr * sign(state) - Reference: - Jeremy Bernstein, Yu-Xiang Wang, Kamyar Azizzadenesheli & Anima Anandkumar. (2018). + Reference: + Jeremy Bernstein, Yu-Xiang Wang, Kamyar Azizzadenesheli & Anima Anandkumar. (2018). signSGD: Compressed Optimisation for Non-Convex Problems. In ICML’18. See: https://arxiv.org/abs/1802.04434 From 98775134d4134e23ff300a472753b4f20a1efda6 Mon Sep 17 00:00:00 2001 From: yuxiangw Date: Tue, 14 Aug 2018 11:12:41 -0700 Subject: [PATCH 3/5] fixing more lint --- python/mxnet/optimizer.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/python/mxnet/optimizer.py b/python/mxnet/optimizer.py index 7be767d52174..ef6a5bb28dd7 100644 --- a/python/mxnet/optimizer.py +++ b/python/mxnet/optimizer.py @@ -548,7 +548,7 @@ def update_multi_precision(self, index, weight, grad, state): @register class Signum(Optimizer): - """The Signum optimizer that takes the sign of gradient or momentum. + r"""The Signum optimizer that takes the sign of gradient or momentum. The optimizer updates the weight by:: From f4489553c133efc8c5a8dea3171770e238c1b948 Mon Sep 17 00:00:00 2001 From: yuxiangw Date: Tue, 14 Aug 2018 18:08:10 -0700 Subject: [PATCH 4/5] jenkins --- python/mxnet/optimizer.py | 1 - 1 file changed, 1 deletion(-) diff --git a/python/mxnet/optimizer.py b/python/mxnet/optimizer.py index ef6a5bb28dd7..622f2b736c3e 100644 --- a/python/mxnet/optimizer.py +++ b/python/mxnet/optimizer.py @@ -15,7 +15,6 @@ # specific language governing permissions and limitations # under the License. -# coding: utf-8 # pylint: disable=too-many-lines """Weight updating functions.""" import logging From 2905ddc75c77675f4acda7d8048a281bbe4845db Mon Sep 17 00:00:00 2001 From: yuxiangw Date: Tue, 14 Aug 2018 22:41:09 -0700 Subject: [PATCH 5/5] adding the coding line utf-8 --- python/mxnet/optimizer.py | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/python/mxnet/optimizer.py b/python/mxnet/optimizer.py index 622f2b736c3e..b69d0c9af0dc 100644 --- a/python/mxnet/optimizer.py +++ b/python/mxnet/optimizer.py @@ -1,3 +1,4 @@ +# coding: utf-8 # Licensed to the Apache Software Foundation (ASF) under one # or more contributor license agreements. See the NOTICE file # distributed with this work for additional information @@ -557,7 +558,7 @@ class Signum(Optimizer): Reference: Jeremy Bernstein, Yu-Xiang Wang, Kamyar Azizzadenesheli & Anima Anandkumar. (2018). - signSGD: Compressed Optimisation for Non-Convex Problems. In ICML’18. + signSGD: Compressed Optimisation for Non-Convex Problems. In ICML'18. See: https://arxiv.org/abs/1802.04434