Update memory_format_tutorial.py (#1925)

mingfeima · Svetlana Karslioglu · web-flow · commit c2575d1b9a26 · 2022-06-27T13:25:19.000-07:00
* Update memory_format_tutorial.py

Add perf gain data for CPU Channels last part.

* Update memory_format_tutorial.py

specify that channels last applies to both cpu and gpu.

Co-authored-by: Svetlana Karslioglu &lt;svekars@fb.com&gt;
diff --git a/intermediate_source/memory_format_tutorial.py b/intermediate_source/memory_format_tutorial.py
@@ -151,7 +151,8 @@
 ######################################################################
 # Performance Gains
 # --------------------------------------------------------------------
-# The most significant performance gains are observed on NVidia's
+# Channels last memory format optimizations are available on both GPU and CPU.
+# On GPU, the most significant performance gains are observed on NVidia's
 # hardware with Tensor Cores support running on reduced precision
 # (``torch.float16``).
 # We were able to archive over 22% perf gains with channels last
@@ -240,6 +241,11 @@
 # ``alexnet``, ``mnasnet0_5``, ``mnasnet0_75``, ``mnasnet1_0``, ``mnasnet1_3``, ``mobilenet_v2``, ``resnet101``, ``resnet152``, ``resnet18``, ``resnet34``, ``resnet50``, ``resnext50_32x4d``, ``shufflenet_v2_x0_5``, ``shufflenet_v2_x1_0``, ``shufflenet_v2_x1_5``, ``shufflenet_v2_x2_0``, ``squeezenet1_0``, ``squeezenet1_1``, ``vgg11``, ``vgg11_bn``, ``vgg13``, ``vgg13_bn``, ``vgg16``, ``vgg16_bn``, ``vgg19``, ``vgg19_bn``, ``wide_resnet101_2``, ``wide_resnet50_2``
 #
 
+######################################################################
+# The following list of models has the full support of Channels last and showing 26%-76% perf gains on Intel(R) Xeon(R) Ice Lake (or newer) CPUs:
+# ``alexnet``, ``densenet121``, ``densenet161``, ``densenet169``, ``googlenet``, ``inception_v3``, ``mnasnet0_5``, ``mnasnet1_0``, ``resnet101``, ``resnet152``, ``resnet18``, ``resnet34``, ``resnet50``, ``resnext101_32x8d``, ``resnext50_32x4d``, ``shufflenet_v2_x0_5``, ``shufflenet_v2_x1_0``, ``squeezenet1_0``, ``squeezenet1_1``, ``vgg11``, ``vgg11_bn``, ``vgg13``, ``vgg13_bn``, ``vgg16``, ``vgg16_bn``, ``vgg19``, ``vgg19_bn``, ``wide_resnet101_2``, ``wide_resnet50_2``
+#
+
 ######################################################################
 # Converting existing models
 # --------------------------
