[doc] Add overflow notice for cuFFT on half precision (#35594)

xwang233 · facebook-github-bot · commit e021c13d2d06 · 2020-03-30T15:53:32.000-07:00
Summary: This would fix #33485. cc ptrblck Pull Request resolved: #35594 Differential Revision: D20722426 Pulled By: ngimel fbshipit-source-id: 68f7304de5d6cecdd9e34e8697fc84bc551b1a45
diff --git a/torch/_torch_docs.py b/torch/_torch_docs.py
@@ -6843,6 +6843,11 @@ def merge_dicts(*dicts):
     configuration. See :ref:`cufft-plan-cache` for more details on how to
     monitor and control the cache.
 
+.. warning::
+    Due to limited dynamic range of half datatype, performing this operation 
+    in half precision may cause the first element of 
+    result to overflow for certain inputs. 
+
 .. warning::
     For CPU tensors, this method is currently only available with MKL. Use
     :func:`torch.backends.mkl.is_available` to check if MKL is installed.
@@ -6938,6 +6943,10 @@ def merge_dicts(*dicts):
     configuration. See :ref:`cufft-plan-cache` for more details on how to
     monitor and control the cache.
 
+.. warning::
+    Due to limited dynamic range of half datatype, performing this operation in half 
+    precision may cause the first element of result to overflow for certain inputs. 
+
 .. warning::
     For CPU tensors, this method is currently only available with MKL. Use
     :func:`torch.backends.mkl.is_available` to check if MKL is installed.
@@ -7022,6 +7031,10 @@ def merge_dicts(*dicts):
     configuration. See :ref:`cufft-plan-cache` for more details on how to
     monitor and control the cache.
 
+.. warning::
+    Due to limited dynamic range of half datatype, performing this operation in half 
+    precision may cause the first element of result to overflow for certain inputs. 
+
 .. warning::
     For CPU tensors, this method is currently only available with MKL. Use
     :func:`torch.backends.mkl.is_available` to check if MKL is installed.
@@ -7099,6 +7112,10 @@ def merge_dicts(*dicts):
     configuration. See :ref:`cufft-plan-cache` for more details on how to
     monitor and control the cache.
 
+.. warning::
+    Due to limited dynamic range of half datatype, performing this operation in half 
+    precision may cause the first element of result to overflow for certain inputs. 
+    
 .. warning::
     For CPU tensors, this method is currently only available with MKL. Use
     :func:`torch.backends.mkl.is_available` to check if MKL is installed.
