OSSForks/pytorch
1
0
Fork 0
mirror of https://github.com/zebrajr/pytorch.git synced 2025-12-06 12:20:52 +01:00
Code Issues Actions 38 Packages Projects Releases Wiki Activity

A bunch of typos (#149404)

Browse Source 
Improves readability.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/149404
Approved by: https://github.com/soulitzer
This commit is contained in:
Dmitry Nikolayev 2025-03-24 16:16:02 +00:00 committed by PyTorch MergeBot
parent ddc0fe903f
commit db92d0f388
1 changed files with 4 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
docs/cpp/source/notes/inference_mode.rst
View File

@ -6,7 +6,7 @@ to be used when you are certain your operations will have no interactions
with autograd (e.g. model training). Compared to ``NoGradMode``, code run with autograd (e.g. model training). Compared to ``NoGradMode``, code run
under this mode gets better performance by disabling autograd related work like under this mode gets better performance by disabling autograd related work like
view tracking and version counter bumps. However, tensors created inside view tracking and version counter bumps. However, tensors created inside
``c10::InferenceMode`` has more limitation when interacting with autograd system as well. ``c10::InferenceMode`` have more limitations when interacting with autograd system as well.
``InferenceMode`` can be enabled for a given block of code. Inside ``InferenceMode`` ``InferenceMode`` can be enabled for a given block of code. Inside ``InferenceMode``
all newly allocated (non-view) tensors are marked as inference tensors. Inference tensors: all newly allocated (non-view) tensors are marked as inference tensors. Inference tensors:
@ -19,7 +19,7 @@ all newly allocated (non-view) tensors are marked as inference tensors. Inferenc
To work around you can make a clone outside ``InferenceMode`` to get a normal tensor before mutating. To work around you can make a clone outside ``InferenceMode`` to get a normal tensor before mutating.
A non-view tensor is an inference tensor if and only if it was allocated inside ``InferenceMode``. A non-view tensor is an inference tensor if and only if it was allocated inside ``InferenceMode``.
A view tensor is an inference tensor if and only if the tensor it is a view of is an inference tensor. A view tensor is an inference tensor if and only if it is a view of an inference tensor.
Inside an ``InferenceMode`` block, we make the following performance guarantees: Inside an ``InferenceMode`` block, we make the following performance guarantees:
@ -66,7 +66,7 @@ users should pay additional attention to:
also affects tensor creation while ``AutoNonVariableTypeMode`` doesn't. In other words, tensors created also affects tensor creation while ``AutoNonVariableTypeMode`` doesn't. In other words, tensors created
inside ``InferenceMode`` are marked as inference tensors so that certain limitation can be applied after inside ``InferenceMode`` are marked as inference tensors so that certain limitation can be applied after
exiting ``InferenceMode``. exiting ``InferenceMode``.
- Enabled/disabled ``InferenceMode`` states can be nested while ``AutoNonVariableTypeMode`` only allows enabled state.. - Enabled/disabled ``InferenceMode`` states can be nested while ``AutoNonVariableTypeMode`` only allows enabled state.
.. code-block:: cpp .. code-block:: cpp
@ -82,7 +82,7 @@ users should pay additional attention to:
// InferenceMode is off // InferenceMode is off
2. Users trying to implement a customized kernel who wants to redispatch under ``Autograd`` dispatch 2. Users trying to implement a customized kernel who want to redispatch under ``Autograd`` dispatch
keys should use ``AutoDispatchBelowADInplaceOrView`` instead. Note ``AutoDispatchBelowADInplaceOrView`` is just a new name keys should use ``AutoDispatchBelowADInplaceOrView`` instead. Note ``AutoDispatchBelowADInplaceOrView`` is just a new name
of ``AutoNonVariableTypeMode`` since it explains the guard's functionality better. We're deprecating of ``AutoNonVariableTypeMode`` since it explains the guard's functionality better. We're deprecating
``AutoNonVariableTypeMode`` and it'll be removed in 1.10 release. See customized kernel ``AutoNonVariableTypeMode`` and it'll be removed in 1.10 release. See customized kernel