Docs update on split semantics.

CHANGELOG.md

@@ -2,8 +2,8 @@
 
 - Update documentation theme to "Read the Docs"
 - [#429](https://github.com/helmholtz-analytics/heat/pull/429) Create submodule for Linear Algebra functions
-- [#429](https://github.com/helmholtz-analytics/heat/pull/429) Implementated QR
-- [#429](https://github.com/helmholtz-analytics/heat/pull/429) Implementated a tiling class to create Square tiles along the diagonal of a 2D matrix
+- [#429](https://github.com/helmholtz-analytics/heat/pull/429) Implemented QR
+- [#429](https://github.com/helmholtz-analytics/heat/pull/429) Implemented a tiling class to create Square tiles along the diagonal of a 2D matrix
 - [#429](https://github.com/helmholtz-analytics/heat/pull/429) Added PyTorch Jitter to inner function of matmul for increased speed
 - [#483](https://github.com/helmholtz-analytics/heat/pull/483) Bugfix: Underlying torch tensor moves to the right device on array initialisation
 - [#483](https://github.com/helmholtz-analytics/heat/pull/483) Bugfix: DNDarray.cpu() changes heat device to cpu

heat/core/manipulations.py

@@ -1196,7 +1196,14 @@ def squeeze(x, axis=None):
     --------
     squeezed : ht.DNDarray
                The input tensor, but with all or a subset of the dimensions of length 1 removed.
-
+               Split semantics: see note below.
+
+    Note:
+    -----
+    Split semantics: if the input tensor is distributed ('x.split is not None'), the output
+    tensor will be distributed along the same dimension. The case where 'x.split = axis'
+    implies a redistribution of the input tensor along the next possible split dimension
+    (cf. 'heat.resplit()') which may affect performance.
 
     Examples:
     >>> import heat as ht