From 19939657eaf5793af62033c34c071c949c409012 Mon Sep 17 00:00:00 2001
From: tensor-tang <jian.j.tang@intel.com>
Date: Thu, 21 Dec 2017 23:47:34 +0800
Subject: [PATCH 1/4] enable training alexnet benchmark

---
 benchmark/paddle/image/alexnet.py            | 22 +++++++++++++++++---
 benchmark/paddle/image/run_mkl_train.sh      |  1 +
 benchmark/paddle/image/run_openblas_train.sh |  1 +
 3 files changed, 21 insertions(+), 3 deletions(-)

diff --git a/benchmark/paddle/image/alexnet.py b/benchmark/paddle/image/alexnet.py
index 3358d43a4b..10db194485 100644
--- a/benchmark/paddle/image/alexnet.py
+++ b/benchmark/paddle/image/alexnet.py
@@ -6,6 +6,7 @@ height = 227
 width = 227
 num_class = 1000
 batch_size = get_config_arg('batch_size', int, 128)
+use_mkldnn = get_config_arg('use_mkldnn', bool, False)
 
 args = {'height': height, 'width': width, 'color': True, 'num_class': num_class}
 define_py_data_sources2(
@@ -31,7 +32,12 @@ net = img_pool_layer(input=net, pool_size=3, stride=2)
 
 # conv2
 net = img_conv_layer(
-    input=net, filter_size=5, num_filters=256, stride=1, padding=2, groups=1)
+    input=net,
+    filter_size=5,
+    num_filters=256,
+    stride=1,
+    padding=2,
+    groups=2 if use_mkldnn else 1)
 net = img_cmrnorm_layer(input=net, size=5, scale=0.0001, power=0.75)
 net = img_pool_layer(input=net, pool_size=3, stride=2)
 
@@ -40,11 +46,21 @@ net = img_conv_layer(
     input=net, filter_size=3, num_filters=384, stride=1, padding=1)
 # conv4
 net = img_conv_layer(
-    input=net, filter_size=3, num_filters=384, stride=1, padding=1, groups=1)
+    input=net,
+    filter_size=3,
+    num_filters=384,
+    stride=1,
+    padding=1,
+    groups=2 if use_mkldnn else 1)
 
 # conv5
 net = img_conv_layer(
-    input=net, filter_size=3, num_filters=256, stride=1, padding=1, groups=1)
+    input=net,
+    filter_size=3,
+    num_filters=256,
+    stride=1,
+    padding=1,
+    groups=2 if use_mkldnn else 1)
 net = img_pool_layer(input=net, pool_size=3, stride=2)
 
 net = fc_layer(
diff --git a/benchmark/paddle/image/run_mkl_train.sh b/benchmark/paddle/image/run_mkl_train.sh
index 5335af5ac1..c38b3e3621 100755
--- a/benchmark/paddle/image/run_mkl_train.sh
+++ b/benchmark/paddle/image/run_mkl_train.sh
@@ -47,5 +47,6 @@ for use_mkldnn in True False; do
     train vgg 19 $batchsize $use_mkldnn
     train resnet 50 $batchsize $use_mkldnn
     train googlenet v1 $batchsize $use_mkldnn
+    train alexnet group2 $batchsize $use_mkldnn
   done
 done
diff --git a/benchmark/paddle/image/run_openblas_train.sh b/benchmark/paddle/image/run_openblas_train.sh
index b9494ce119..caea5548c3 100755
--- a/benchmark/paddle/image/run_openblas_train.sh
+++ b/benchmark/paddle/image/run_openblas_train.sh
@@ -36,4 +36,5 @@ for batchsize in 64 128 256; do
   train vgg 19 $batchsize
   train resnet 50 $batchsize
   train googlenet v1 $batchsize
+  train alexnet group2 $batchsize $use_mkldnn
 done

From 86b8bdc0af01960eaded403ade4214faebe5c475 Mon Sep 17 00:00:00 2001
From: tensor-tang <jian.j.tang@intel.com>
Date: Thu, 21 Dec 2017 23:54:38 +0800
Subject: [PATCH 2/4] enable inference alexnet benchmark

---
 benchmark/paddle/image/alexnet.py            | 11 ++++++++++-
 benchmark/paddle/image/run_mkl_infer.sh      |  1 +
 benchmark/paddle/image/run_openblas_infer.sh |  1 +
 3 files changed, 12 insertions(+), 1 deletion(-)

diff --git a/benchmark/paddle/image/alexnet.py b/benchmark/paddle/image/alexnet.py
index 10db194485..b0beef8ca7 100644
--- a/benchmark/paddle/image/alexnet.py
+++ b/benchmark/paddle/image/alexnet.py
@@ -7,8 +7,17 @@ width = 227
 num_class = 1000
 batch_size = get_config_arg('batch_size', int, 128)
 use_mkldnn = get_config_arg('use_mkldnn', bool, False)
+is_infer = get_config_arg("is_infer", bool, False)
+num_samples = get_config_arg('num_samples', int, 2560)
 
-args = {'height': height, 'width': width, 'color': True, 'num_class': num_class}
+args = {
+    'height': height,
+    'width': width,
+    'color': True,
+    'num_class': num_class,
+    'is_infer': is_infer,
+    'num_samples': num_samples
+}
 define_py_data_sources2(
     "train.list", None, module="provider", obj="process", args=args)
 
diff --git a/benchmark/paddle/image/run_mkl_infer.sh b/benchmark/paddle/image/run_mkl_infer.sh
index d795bcab1b..00942e32a5 100755
--- a/benchmark/paddle/image/run_mkl_infer.sh
+++ b/benchmark/paddle/image/run_mkl_infer.sh
@@ -79,6 +79,7 @@ fi
 # inference benchmark
 for use_mkldnn in True False; do
   for batchsize in 1 2 4 8 16; do
+    infer alexnet group2 $batchsize $use_mkldnn
     infer googlenet v1 $batchsize $use_mkldnn
     infer resnet 50 $batchsize $use_mkldnn
     infer vgg 19 $batchsize $use_mkldnn
diff --git a/benchmark/paddle/image/run_openblas_infer.sh b/benchmark/paddle/image/run_openblas_infer.sh
index c1001d3a7c..3dad42ee0d 100755
--- a/benchmark/paddle/image/run_openblas_infer.sh
+++ b/benchmark/paddle/image/run_openblas_infer.sh
@@ -56,6 +56,7 @@ fi
 
 # inference benchmark
 for batchsize in 1 2 4 8 16; do
+  infer alexnet group2 $batchsize $use_mkldnn
   infer googlenet v1 $batchsize
   infer resnet 50 $batchsize
   infer vgg 19 $batchsize

From a785496b6904fe56754110242a8dceb8ef795221 Mon Sep 17 00:00:00 2001
From: tensor-tang <jian.j.tang@intel.com>
Date: Fri, 22 Dec 2017 10:19:33 +0800
Subject: [PATCH 3/4] fix logical error

---
 benchmark/paddle/image/alexnet.py            | 32 ++++++--------------
 benchmark/paddle/image/run_mkl_infer.sh      |  2 +-
 benchmark/paddle/image/run_mkl_train.sh      |  2 +-
 benchmark/paddle/image/run_openblas_infer.sh |  2 +-
 benchmark/paddle/image/run_openblas_train.sh |  2 +-
 5 files changed, 14 insertions(+), 26 deletions(-)

diff --git a/benchmark/paddle/image/alexnet.py b/benchmark/paddle/image/alexnet.py
index b0beef8ca7..77d130ae34 100644
--- a/benchmark/paddle/image/alexnet.py
+++ b/benchmark/paddle/image/alexnet.py
@@ -6,7 +6,7 @@ height = 227
 width = 227
 num_class = 1000
 batch_size = get_config_arg('batch_size', int, 128)
-use_mkldnn = get_config_arg('use_mkldnn', bool, False)
+gp = get_config_arg('layer_num', int, 1)
 is_infer = get_config_arg("is_infer", bool, False)
 num_samples = get_config_arg('num_samples', int, 2560)
 
@@ -41,12 +41,7 @@ net = img_pool_layer(input=net, pool_size=3, stride=2)
 
 # conv2
 net = img_conv_layer(
-    input=net,
-    filter_size=5,
-    num_filters=256,
-    stride=1,
-    padding=2,
-    groups=2 if use_mkldnn else 1)
+    input=net, filter_size=5, num_filters=256, stride=1, padding=2, groups=gp)
 net = img_cmrnorm_layer(input=net, size=5, scale=0.0001, power=0.75)
 net = img_pool_layer(input=net, pool_size=3, stride=2)
 
@@ -55,21 +50,11 @@ net = img_conv_layer(
     input=net, filter_size=3, num_filters=384, stride=1, padding=1)
 # conv4
 net = img_conv_layer(
-    input=net,
-    filter_size=3,
-    num_filters=384,
-    stride=1,
-    padding=1,
-    groups=2 if use_mkldnn else 1)
+    input=net, filter_size=3, num_filters=384, stride=1, padding=1, groups=gp)
 
 # conv5
 net = img_conv_layer(
-    input=net,
-    filter_size=3,
-    num_filters=256,
-    stride=1,
-    padding=1,
-    groups=2 if use_mkldnn else 1)
+    input=net, filter_size=3, num_filters=256, stride=1, padding=1, groups=gp)
 net = img_pool_layer(input=net, pool_size=3, stride=2)
 
 net = fc_layer(
@@ -84,6 +69,9 @@ net = fc_layer(
     layer_attr=ExtraAttr(drop_rate=0.5))
 net = fc_layer(input=net, size=1000, act=SoftmaxActivation())
 
-lab = data_layer('label', num_class)
-loss = cross_entropy(input=net, label=lab)
-outputs(loss)
+if is_infer:
+    outputs(net)
+else:
+    lab = data_layer('label', num_class)
+    loss = cross_entropy(input=net, label=lab)
+    outputs(loss)
diff --git a/benchmark/paddle/image/run_mkl_infer.sh b/benchmark/paddle/image/run_mkl_infer.sh
index 00942e32a5..a3b5e2db5e 100755
--- a/benchmark/paddle/image/run_mkl_infer.sh
+++ b/benchmark/paddle/image/run_mkl_infer.sh
@@ -79,7 +79,7 @@ fi
 # inference benchmark
 for use_mkldnn in True False; do
   for batchsize in 1 2 4 8 16; do
-    infer alexnet group2 $batchsize $use_mkldnn
+    infer alexnet 2 $batchsize $use_mkldnn
     infer googlenet v1 $batchsize $use_mkldnn
     infer resnet 50 $batchsize $use_mkldnn
     infer vgg 19 $batchsize $use_mkldnn
diff --git a/benchmark/paddle/image/run_mkl_train.sh b/benchmark/paddle/image/run_mkl_train.sh
index c38b3e3621..03d2d378fb 100755
--- a/benchmark/paddle/image/run_mkl_train.sh
+++ b/benchmark/paddle/image/run_mkl_train.sh
@@ -47,6 +47,6 @@ for use_mkldnn in True False; do
     train vgg 19 $batchsize $use_mkldnn
     train resnet 50 $batchsize $use_mkldnn
     train googlenet v1 $batchsize $use_mkldnn
-    train alexnet group2 $batchsize $use_mkldnn
+    train alexnet 2 $batchsize $use_mkldnn
   done
 done
diff --git a/benchmark/paddle/image/run_openblas_infer.sh b/benchmark/paddle/image/run_openblas_infer.sh
index 3dad42ee0d..ec9235e2c2 100755
--- a/benchmark/paddle/image/run_openblas_infer.sh
+++ b/benchmark/paddle/image/run_openblas_infer.sh
@@ -56,7 +56,7 @@ fi
 
 # inference benchmark
 for batchsize in 1 2 4 8 16; do
-  infer alexnet group2 $batchsize $use_mkldnn
+  infer alexnet 2 $batchsize $use_mkldnn
   infer googlenet v1 $batchsize
   infer resnet 50 $batchsize
   infer vgg 19 $batchsize
diff --git a/benchmark/paddle/image/run_openblas_train.sh b/benchmark/paddle/image/run_openblas_train.sh
index caea5548c3..1e007be966 100755
--- a/benchmark/paddle/image/run_openblas_train.sh
+++ b/benchmark/paddle/image/run_openblas_train.sh
@@ -36,5 +36,5 @@ for batchsize in 64 128 256; do
   train vgg 19 $batchsize
   train resnet 50 $batchsize
   train googlenet v1 $batchsize
-  train alexnet group2 $batchsize $use_mkldnn
+  train alexnet 2 $batchsize $use_mkldnn
 done

From 025a6f3c234c07ac34d881db7f9f4dbb47be25b4 Mon Sep 17 00:00:00 2001
From: tensor-tang <jian.j.tang@intel.com>
Date: Fri, 22 Dec 2017 10:33:02 +0800
Subject: [PATCH 4/4] unify the test reorder

---
 benchmark/paddle/image/run_mkl_infer.sh      | 6 +++---
 benchmark/paddle/image/run_openblas_infer.sh | 6 +++---
 benchmark/paddle/image/run_openblas_train.sh | 2 +-
 3 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/benchmark/paddle/image/run_mkl_infer.sh b/benchmark/paddle/image/run_mkl_infer.sh
index a3b5e2db5e..c22c4deb1c 100755
--- a/benchmark/paddle/image/run_mkl_infer.sh
+++ b/benchmark/paddle/image/run_mkl_infer.sh
@@ -79,9 +79,9 @@ fi
 # inference benchmark
 for use_mkldnn in True False; do
   for batchsize in 1 2 4 8 16; do
-    infer alexnet 2 $batchsize $use_mkldnn
-    infer googlenet v1 $batchsize $use_mkldnn
-    infer resnet 50 $batchsize $use_mkldnn
     infer vgg 19 $batchsize $use_mkldnn
+    infer resnet 50 $batchsize $use_mkldnn
+    infer googlenet v1 $batchsize $use_mkldnn
+    infer alexnet 2 $batchsize $use_mkldnn
   done
 done
diff --git a/benchmark/paddle/image/run_openblas_infer.sh b/benchmark/paddle/image/run_openblas_infer.sh
index ec9235e2c2..ba9019c9de 100755
--- a/benchmark/paddle/image/run_openblas_infer.sh
+++ b/benchmark/paddle/image/run_openblas_infer.sh
@@ -56,8 +56,8 @@ fi
 
 # inference benchmark
 for batchsize in 1 2 4 8 16; do
-  infer alexnet 2 $batchsize $use_mkldnn
-  infer googlenet v1 $batchsize
-  infer resnet 50 $batchsize
   infer vgg 19 $batchsize
+  infer resnet 50 $batchsize 
+  infer googlenet v1 $batchsize
+  infer alexnet 2 $batchsize
 done
diff --git a/benchmark/paddle/image/run_openblas_train.sh b/benchmark/paddle/image/run_openblas_train.sh
index 1e007be966..a1b5ee9da8 100755
--- a/benchmark/paddle/image/run_openblas_train.sh
+++ b/benchmark/paddle/image/run_openblas_train.sh
@@ -36,5 +36,5 @@ for batchsize in 64 128 256; do
   train vgg 19 $batchsize
   train resnet 50 $batchsize
   train googlenet v1 $batchsize
-  train alexnet 2 $batchsize $use_mkldnn
+  train alexnet 2 $batchsize
 done