Merge branch 'master' into ptddp-launcher

Author: vishwakaria

CHANGELOG.md

@@ -1,5 +1,28 @@
 # Changelog
 
+## v2.100.0 (2022-07-18)
+
+### Features
+
+ * upgrade to support python 3.10
+ * Add target_model to support multi-model endpoints
+ * Added support for feature group schema change and feature parameters
+
+### Bug Fixes and Other Changes
+
+ * enable model.register without 'inference' & 'transform' instances
+ * rename RegisterModel inner steps to prevent duplicate step names
+ * remove primitive_or_expr() from conditions
+ * support pipeline variables for spark processors run arguments
+ * make 'ModelInput' field optional for inference recommendation
+ * Fix processing image uri param
+ * fix: neo inferentia as compilation target not using framework ver
+
+### Documentation Changes
+
+ * SageMaker model parallel library v1.10.0 documentation
+ * add detail & links to clarify docstrings
+
 ## v2.99.0 (2022-07-08)
 
 ### Features

VERSION

@@ -1 +1 @@
-2.99.1.dev0
+2.100.1.dev0

doc/api/training/smd_model_parallel_general.rst

@@ -178,6 +178,16 @@ PyTorch-specific Parameters
     - 1
     - The number of devices over which the tensor parallel modules will be distributed.
       If ``tensor_parallel_degree`` is greater than 1, then ``ddp`` must be set to ``True``.
+  * - ``fp16`` (**smdistributed-modelparallel**>=v1.10)
+    - bool
+    - ``False``
+    - To run FP16 training, add ``"fp16"'": True`` to the smp configuration.
+      Other APIs remain the same between FP16 and FP32.
+      If ``fp16`` is enabled and when user calls ``smp.DistributedModel``,
+      the model will be wrapped with ``FP16_Module``, which converts the model
+      to FP16 dtype and deals with forward pass in FP16.
+      If ``fp16`` is enabled and when user calls ``smp.DistributedOptimizer``,
+      the optimizer will be wrapped with ``FP16_Optimizer``.
   * - ``fp16_params`` (**smdistributed-modelparallel**>=v1.6)
     - bool
     - ``False``

doc/api/training/smp_versions/archives.rst

@@ -3,6 +3,7 @@
 .. toctree::
     :maxdepth: 1
 
+    v1_9_0.rst
     v1_6_0.rst
     v1_5_0.rst
     v1_4_0.rst

doc/api/training/smp_versions/latest.rst

@@ -10,7 +10,7 @@ depending on which version of the library you need to use.
 To use the library, reference the
 **Common API** documentation alongside the framework specific API documentation.
 
-Version 1.7.0, 1.8.0, 1.8.1, 1.9.0 (Latest)
+Version 1.10.0 (Latest)
 ===========================================
 
 To use the library, reference the Common API documentation alongside the framework specific API documentation.