diff --git a/docs/src/learn/reference/index.rst b/docs/src/learn/reference/index.rst
index 9db804019..43a977c50 100644
--- a/docs/src/learn/reference/index.rst
+++ b/docs/src/learn/reference/index.rst
@@ -12,6 +12,10 @@ API reference
   .. grid::
     :margin: 0 0 0 0
 
+    .. grid-item-version:: 0.2.2
+        :tag-prefix: metatensor-learn-v
+        :url-suffix: learn/reference/index.html
+
     .. grid-item-version:: 0.2.1
         :tag-prefix: metatensor-learn-v
         :url-suffix: learn/reference/index.html
diff --git a/python/metatensor-learn/CHANGELOG.md b/python/metatensor-learn/CHANGELOG.md
index 77a64b9cc..fd9e9e4ee 100644
--- a/python/metatensor-learn/CHANGELOG.md
+++ b/python/metatensor-learn/CHANGELOG.md
@@ -17,6 +17,8 @@ follows [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 ### Removed
 -->
 
+## [Version 0.2.2](https://github.com/lab-cosmo/metatensor/releases/tag/metatensor-learn-v0.2.2) - 2024-05-16
+
 ### Added
 
 - Added torch-style activation function module maps to `metatensor.learn.nn`: `ReLU`,
@@ -26,6 +28,7 @@ follows [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
   and `InvariantTanh` (#513)
 
 ### Fixed
+
 - Set correct device for output of when torch default device is different than input device (#595)
 
 ## [Version 0.2.1](https://github.com/lab-cosmo/metatensor/releases/tag/metatensor-learn-v0.2.1) - 2024-03-01
diff --git a/python/metatensor-learn/setup.py b/python/metatensor-learn/setup.py
index 73f2100ed..b2968026a 100644
--- a/python/metatensor-learn/setup.py
+++ b/python/metatensor-learn/setup.py
@@ -14,7 +14,7 @@
     os.path.join(ROOT, "..", "metatensor-operations")
 )
 
-METATENSOR_LEARN_VERSION = "0.2.1"
+METATENSOR_LEARN_VERSION = "0.2.2"
 
 
 class bdist_egg_disabled(bdist_egg):