mirror of
https://github.com/pytorch/pytorch.git
synced 2025-10-20 21:14:14 +08:00
Fix typo under docs directory and RELEASE.md (#85896)
This PR fixes typo in rst files under docs directory and `RELEASE.md`. Pull Request resolved: https://github.com/pytorch/pytorch/pull/85896 Approved by: https://github.com/kit1980
This commit is contained in:
Kazuaki Ishizaki
committed by
PyTorch MergeBot
PyTorch MergeBot
parent
11224f34b8
commit
bc57306bdd
@ -14,7 +14,7 @@
|
||||
- [Release Candidate health validation](#release-candidate-health-validation)
|
||||
- [Cherry Picking Fixes](#cherry-picking-fixes)
|
||||
- [Promoting RCs to Stable](#promoting-rcs-to-stable)
|
||||
- [Additonal Steps to prepare for release day](#additonal-steps-to-prepare-for-release-day)
|
||||
- [Additional Steps to prepare for release day](#additional-steps-to-prepare-for-release-day)
|
||||
- [Modify release matrix](#modify-release-matrix)
|
||||
- [Open Google Colab issue](#open-google-colab-issue)
|
||||
- [Patch Releases](#patch-releases)
|
||||
@ -186,7 +186,7 @@ Promotion should occur in two steps:
|
||||
|
||||
**NOTE**: The promotion of wheels to PyPI can only be done once so take caution when attempting to promote wheels to PyPI, (see https://github.com/pypa/warehouse/issues/726 for a discussion on potential draft releases within PyPI)
|
||||
|
||||
## Additonal Steps to prepare for release day
|
||||
## Additional Steps to prepare for release day
|
||||
|
||||
The following should be prepared for the release day
|
||||
|
||||
@ -264,7 +264,7 @@ For versions of Python that we support we follow the [NEP 29 policy](https://num
|
||||
|
||||
## Accelerator Software
|
||||
|
||||
For acclerator software like CUDA and ROCm we will typically use the following criteria:
|
||||
For accelerator software like CUDA and ROCm we will typically use the following criteria:
|
||||
* Support latest 2 minor versions
|
||||
|
||||
### Special support cases
|
||||
|
||||
@ -61,7 +61,7 @@ Pytorch's C++ API provides the following ways to set CUDA stream:
|
||||
|
||||
.. attention::
|
||||
|
||||
This function may have nosthing to do with the current device. It only changes the current stream on the stream's device.
|
||||
This function may have nothing to do with the current device. It only changes the current stream on the stream's device.
|
||||
We recommend using ``CUDAStreamGuard``, instead, since it switches to the stream's device and makes it the current stream on that device.
|
||||
``CUDAStreamGuard`` will also restore the current device and stream when it's destroyed
|
||||
|
||||
|
||||
@ -203,7 +203,7 @@ grad mode in the next forward pass.
|
||||
|
||||
The implementations in :ref:`nn-init-doc` also
|
||||
rely on no-grad mode when initializing the parameters as to avoid
|
||||
autograd tracking when updating the intialized parameters in-place.
|
||||
autograd tracking when updating the initialized parameters in-place.
|
||||
|
||||
Inference Mode
|
||||
^^^^^^^^^^^^^^
|
||||
|
||||
@ -543,7 +543,7 @@ as follows:
|
||||
|
||||
where :math:`\text{clamp}(.)` is the same as :func:`~torch.clamp` while the
|
||||
scale :math:`s` and zero point :math:`z` are then computed
|
||||
as decribed in :class:`~torch.ao.quantization.observer.MinMaxObserver`, specifically:
|
||||
as described in :class:`~torch.ao.quantization.observer.MinMaxObserver`, specifically:
|
||||
|
||||
.. math::
|
||||
|
||||
|
||||
@ -80,7 +80,7 @@ The following table compares the differences between Eager Mode Quantization and
|
||||
| |Static, Dynamic, |Static, Dynamic, |
|
||||
| |Weight Only |Weight Only |
|
||||
| | | |
|
||||
| |Quantiztion Aware |Quantiztion Aware |
|
||||
| |Quantization Aware |Quantization Aware |
|
||||
| |Training: |Training: |
|
||||
| |Static |Static |
|
||||
+-----------------+-------------------+-------------------+
|
||||
@ -632,7 +632,7 @@ Quantization Mode Support
|
||||
| |Quantization |Dataset | Works Best For | Accuracy | Notes |
|
||||
| |Mode |Requirement | | | |
|
||||
+-----------------------------+---------------------------------+--------------------+----------------+----------------+------------+-----------------+
|
||||
|Post Training Quantization |Dyanmic/Weight Only Quantization |activation |None |LSTM, MLP, |good |Easy to use, |
|
||||
|Post Training Quantization |Dynamic/Weight Only Quantization |activation |None |LSTM, MLP, |good |Easy to use, |
|
||||
| | |dynamically | |Embedding, | |close to static |
|
||||
| | |quantized (fp16, | |Transformer | |quantization when|
|
||||
| | |int8) or not | | | |performance is |
|
||||
@ -640,7 +640,7 @@ Quantization Mode Support
|
||||
| | |statically quantized| | | |bound due to |
|
||||
| | |(fp16, int8, in4) | | | |weights |
|
||||
| +---------------------------------+--------------------+----------------+----------------+------------+-----------------+
|
||||
| |Static Quantization |acivation and |calibration |CNN |good |Provides best |
|
||||
| |Static Quantization |activation and |calibration |CNN |good |Provides best |
|
||||
| | |weights statically |dataset | | |perf, may have |
|
||||
| | |quantized (int8) | | | |big impact on |
|
||||
| | | | | | |accuracy, good |
|
||||
@ -652,7 +652,7 @@ Quantization Mode Support
|
||||
| | |weight are fake |dataset | | |for now |
|
||||
| | |quantized | | | | |
|
||||
| +---------------------------------+--------------------+----------------+----------------+------------+-----------------+
|
||||
| |Static Quantization |activatio nand |fine-tuning |CNN, MLP, |best |Typically used |
|
||||
| |Static Quantization |activation and |fine-tuning |CNN, MLP, |best |Typically used |
|
||||
| | |weight are fake |dataset |Embedding | |when static |
|
||||
| | |quantized | | | |quantization |
|
||||
| | | | | | |leads to bad |
|
||||
@ -736,7 +736,7 @@ Backend/Hardware Support
|
||||
+-----------------+---------------+------------+------------+------------+
|
||||
|server GPU |TensorRT (early|Not support |Supported |Static |
|
||||
| |prototype) |this it | |Quantization|
|
||||
| | |requries a | | |
|
||||
| | |requires a | | |
|
||||
| | |graph | | |
|
||||
+-----------------+---------------+------------+------------+------------+
|
||||
|
||||
|
||||
@ -16,7 +16,7 @@ machines.
|
||||
CUDA support was introduced in PyTorch 1.9 and is still a **beta** feature.
|
||||
Not all features of the RPC package are yet compatible with CUDA support and
|
||||
thus their use is discouraged. These unsupported features include: RRefs,
|
||||
JIT compatibility, dist autograd and dist optimizier, and profiling. These
|
||||
JIT compatibility, dist autograd and dist optimizer, and profiling. These
|
||||
shortcomings will be addressed in future releases.
|
||||
|
||||
.. note ::
|
||||
|
||||
@ -470,7 +470,7 @@ ncols, *densesize)`` where ``len(batchsize) == B`` and
|
||||
|
||||
The batches of sparse CSR tensors are dependent: the number of
|
||||
specified elements in all batches must be the same. This somewhat
|
||||
artifical constraint allows efficient storage of the indices of
|
||||
artificial constraint allows efficient storage of the indices of
|
||||
different CSR batches.
|
||||
|
||||
.. note::
|
||||
|
||||
Reference in New Issue
Block a user