Neuron SDK 2.18.0 updates (#1618) (#857)

Release notes updates for Neuron SDK 2.18.0

Author: aws-rxgupta

compiler/neuronx-cc/api-reference-guide/neuron-compiler-cli-reference-guide.rst

@@ -60,6 +60,7 @@ Available Commands:
      [--auto-cast-type <data_type>]
      [--distribution-strategy <distribution_type>]
      [--optlevel <opt-level>], or [-O <opt-level>]
+     [--enable-mixed-precision-accumulation]
      [--enable-saturate-infinity]
      [--enable-fast-context-switch>]
      [--enable-fast-loading-neuron-binaries]
@@ -135,7 +136,9 @@ Available Commands:
 
     .. note:: This option supercedes, and deprecates, the ``—enable-experimental-O1`` option introduced in an earlier release.
 
-  - :option:`--enable-saturate-infinity`: Convert +/- infinity values to MAX/MIN_FLOAT for certain computations that have a high risk of generating Not-a-Number (NaN) values. There is a potential performance impact during model execution when this conversion is enabled.
+  - :option:`--enable-mixed-precision-accumulation`: Perform intermediate calculations of accumulation operators (such as softmax and layernorm) in FP32 and cast the result to the model-designated datatype. This improves the operator's resulting accuracy.
+
+  - :option:`--enable-saturate-infinity`: Convert +/- infinity values to MAX/MIN_FLOAT for compiler-introduced matrix-multiply transpose computations that have a high risk of generating Not-a-Number (NaN) values. There is a potential performance impact during model execution when this conversion is enabled.
 
   - :option:`--enable-fast-context-switch`: Optimize for faster model switching rather than execution latency.
       This option will defer loading some weight constants until the start of model execution. This results in overall faster system performance when your application switches between models frequently on the same Neuron Core (or set of cores).

conf.py

@@ -142,7 +142,7 @@
 #top_banner_message="<span>&#9888;</span><a class='reference internal' style='color:white;' href='https://awsdocs-neuron.readthedocs-hosted.com/en/latest/general/setup/setup-troubleshooting.html#gpg-key-update'>  Neuron repository GPG key for Ubuntu installation has expired, see instructions how to update! </a>"
 
 
-top_banner_message="Neuron 2.17.0 is released! check <a class='reference internal' style='color:white;' href='https://awsdocs-neuron.readthedocs-hosted.com/en/latest/release-notes/index.html#latest-neuron-release'> What's New  </a> and <a class='reference internal' style='color:white;' href='https://awsdocs-neuron.readthedocs-hosted.com/en/latest/general/announcements/index.html'> Announcements  </a>"
+top_banner_message="Neuron 2.18.0 is released! check <a class='reference internal' style='color:white;' href='https://awsdocs-neuron.readthedocs-hosted.com/en/latest/release-notes/index.html#latest-neuron-release'> What's New  </a> and <a class='reference internal' style='color:white;' href='https://awsdocs-neuron.readthedocs-hosted.com/en/latest/general/announcements/index.html'> Announcements  </a>"
 
 
 html_theme = "sphinx_book_theme"

containers/developerflows.rst

@@ -9,6 +9,7 @@ Containers - Developer Flows
     /containers/dlc-then-ecs-devflow
     /containers/dlc-then-eks-devflow
     /containers/container-sm-hosting-devflow
+    /containers/dlc-then-customize-devflow
 
 
 

containers/developerflows.txt

@@ -6,3 +6,4 @@
         * :ref:`containers-dlc-then-ecs-devflow`
         * :ref:`containers-dlc-then-eks-devflow`
         * :ref:`containers-byoc-hosting-devflow`
+        * :ref:`containers-dlc-then-customize-devflow`

containers/dlc-then-customize-devflow.rst

@@ -0,0 +1,3 @@
+.. _containers-dlc-then-customize-devflow:
+
+.. include:: /general/devflows/dlc-then-customize-devflow.rst

containers/locate-neuron-dlc-image.txt

@@ -1,12 +1,26 @@
-.. tab-set:: 
+.. tab-set::
 
-    .. tab-item:: Latest Neuron DLC images
+    .. tab-item:: Introduction
 
+        The Pytorch Neuron DLC images are published to ECR Public, which is the recommended URL to use for most cases. If you are working within AWS SageMaker, you should use the Amazon ECR URL instead of the Amazon ECR Public one because of the restriction of Sagemaker. TensorFlow DLCs are not updated with the latest release. For earlier releases please check `here <https://github.com/aws/deep-learning-containers/blob/master/available_images.md#neuron-containers>`_.
+
+.. tab-set::
+
+    .. tab-item:: Neuron DLC images in Amazon ECR Public
+
+        .. df-table::
+         :header-rows: 1
+
+         df = pd.read_csv('neuron_dlc_images.csv')
+
+.. tab-set::
+
+    .. tab-item:: Latest Neuron DLC images in Amazon ECR
 
         Find latest `Neuron DLC images <https://github.com/aws/deep-learning-containers/blob/master/available_images.md#user-content-neuron-containers>`_.
 
-.. tab-set:: 
+.. tab-set::
 
-    .. tab-item:: Locate specific Neuron DLC release
+    .. tab-item:: Locate specific Neuron DLC release in Amazon ECR
 
         In the `DLC release page <https://github.com/aws/deep-learning-containers/releases>`_ do a search for Neuron to get the ECR repo location of specific Neuron DLC release.

containers/neuron_dlc_images.csv

@@ -0,0 +1,6 @@
+Framework,Neuron Package,Job Type,Supported EC2 Instance Types,Python Version Options,ECR Public Repo URL,Image Details,Other Packages
+PyTorch 2.1.2,"aws-neuronx-tools, neuronx_distributed, torch-neuronx, transformers-neuronx",inference,trn1 and inf2,3.10 (py310),https://gallery.ecr.aws/neuron/pytorch-inference-neuronx,https://github.com/aws-neuron/deep-learning-containers#pytorch-inference-neuronx,torchserve
+PyTorch 2.1.2,"aws-neuronx-tools, neuronx_distributed, torch-neuronx",training,trn1 and inf2,3.10 (py310),https://gallery.ecr.aws/neuron/pytorch-training-neuronx,https://github.com/aws-neuron/deep-learning-containers#pytorch-training-neuronx,
+PyTorch 1.13.1,"aws-neuronx-tools, torch-neuron",inference,inf1,3.10 (py310),https://gallery.ecr.aws/neuron/pytorch-inference-neuron,https://github.com/aws-neuron/deep-learning-containers#pytorch-inference-neuron,torchserve
+PyTorch 1.13.1,"aws-neuronx-tools, neuronx_distributed, torch-neuronx, transformers-neuronx",inference,trn1 and inf2,3.10 (py310),https://gallery.ecr.aws/neuron/pytorch-inference-neuronx,https://github.com/aws-neuron/deep-learning-containers#pytorch-inference-neuronx,torchserve
+PyTorch 1.13.1,"aws-neuronx-tools, neuronx_distributed, torch-neuronx",training,trn1 and inf2,3.10 (py310),https://gallery.ecr.aws/neuron/pytorch-training-neuronx,https://github.com/aws-neuron/deep-learning-containers#pytorch-training-neuronx,