[chore] 0.3.9 release (#750)

* [chore] 0.3.9 release

* update changelog

* address comments
Co-authored-by: Min Xu <min.xu.public@gmail.com>

CHANGELOG.md

@@ -6,11 +6,19 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## NEXT - TBD
 ### Fixed
-- FSDP: fixed metadata saving and shard consolidation for MoE cases [#746]
+
+### Added
+
+## [0.3.9] - 2021-07-26
+### Fixed
+- FSDP: fixed metadata saving and shard consolidation for MoE cases. When a model has
+        shared parameters or mixture of expert layers, the handling of state dict
+        metadata was broken. This release fixes that. [#746]
 - OSS: fixed the buckets which would stay in fp16 if `broadcast fp16` was required (#751)
 
 ### Added
-- FSDP: better performance; use `_allgather_base` and `_reduce_scatter_base` when available [#729]
+- FSDP: better performance; use `_allgather_base` and `_reduce_scatter_base` when they are
+        available from pytorch nightly version (will be in 1.10 releases) [#729]
 - FSDP: prepared FSDP internals for supporting multiple groups of flatten parameters (to support more general optimization) [#746]
 
 ## [0.3.8] - 2021-07-12

docs/source/conf.py

@@ -29,7 +29,7 @@ copyright = "2020-2021, Facebook AI Research"
 author = "Facebook AI Research"
 
 # The full version, including alpha/beta/rc tags
-release = "0.3.8"
+release = "0.3.9"
 
 
 # -- General configuration ---------------------------------------------------

fairscale/__init__.py

@@ -4,7 +4,7 @@
 # LICENSE file in the root directory of this source tree.
 
 # Please update the doc version in docs/source/conf.py as well.
-__version__ = "0.3.8"
+__version__ = "0.3.9"
 
 ################################################################################
 # Import most common subpackages