Update the docs of torch.eig about derivative (#47598)

mfkasim1 · facebook-github-bot · commit a49367e9c9fc · 2020-11-09T13:28:27.000-08:00
Summary: Related: #33090 I just realized that I haven't updated the docs of `torch.eig` when implementing the backward. Here's the PR updating the docs about the grad of `torch.eig`. cc albanD Pull Request resolved: #47598 Reviewed By: heitorschueroff Differential Revision: D24829373 Pulled By: albanD fbshipit-source-id: 89963ce66b2933e6c34e2efc93ad0f2c3dd28c68
diff --git a/torch/_torch_docs.py b/torch/_torch_docs.py
@@ -2640,7 +2640,7 @@ def merge_dicts(*dicts):
 differently than dot(a, b). If the first argument is complex, the complex conjugate of the
 first argument is used for the calculation of the dot product.
 
-.. note:: 
+.. note::
 
     Unlike NumPy's vdot, torch.vdot intentionally only supports computing the dot product
     of two 1D tensors with the same number of elements.
@@ -2672,7 +2672,7 @@ def merge_dicts(*dicts):
 
 .. note::
     Since eigenvalues and eigenvectors might be complex, backward pass is supported only
-    for :func:`torch.symeig`
+    if eigenvalues and eigenvectors are all real valued.
 
 Args:
     input (Tensor): the square matrix of shape :math:`(n \times n)` for which the eigenvalues and eigenvectors
