frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-10-20 12:54:11 +08:00
Code Issues Packages Projects Releases Wiki Activity

Fix/issue #155027 (#155252)

Browse Source 
Fixes #155027
Converted RST files to Markdown

Pull Request resolved: https://github.com/pytorch/pytorch/pull/155252
Approved by: https://github.com/svekars

Co-authored-by: Svetlana Karslioglu <svekars@meta.com>
This commit is contained in:
Abhinav Tharamel
2025-06-08 21:17:31 +00:00
committed by PyTorch MergeBot
parent 3d82a1dfb5
commit d41f62b7a0
5 changed files with 319 additions and 266 deletions
Download Patch File Download Diff File Expand all files Collapse all files

38
docs/source/mps.rst → docs/source/mps.md
View File

@ -1,8 +1,14 @@
torch.mps
===================================
.. automodule:: torch.mps
.. currentmodule:: torch.mps
# torch.mps
```{eval-rst}
.. automodule:: torch.mps
```
```{eval-rst}
.. currentmodule:: torch.mps
```
```{eval-rst}
.. autosummary::
:toctree: generated
:nosignatures:
@ -19,9 +25,11 @@ torch.mps
driver_allocated_memory
recommended_max_memory
compile_shader
```
MPS Profiler
------------
## MPS Profiler
```{eval-rst}
.. autosummary::
:toctree: generated
:nosignatures:
@ -33,17 +41,27 @@ MPS Profiler
profiler.is_capturing_metal
profiler.is_metal_capture_enabled
profiler.metal_capture
```
MPS Event
------------
## MPS Event
```{eval-rst}
.. autosummary::
:toctree: generated
:nosignatures:
event.Event
```
.. This module needs to be documented. Adding here in the meantime
.. for tracking purposes
% This module needs to be documented. Adding here in the meantime
% for tracking purposes
```{eval-rst}
.. py:module:: torch.mps.event
```
```{eval-rst}
.. py:module:: torch.mps.profiler
```

18
docs/source/mtia.rst → docs/source/mtia.md
View File

@ -1,11 +1,16 @@
torch.mtia
===================================
# torch.mtia
The MTIA backend is implemented out of the tree, only interfaces are be defined here.
```{eval-rst}
.. automodule:: torch.mtia
.. currentmodule:: torch.mtia
```
```{eval-rst}
.. currentmodule:: torch.mtia
```
```{eval-rst}
.. autosummary::
:toctree: generated
:nosignatures:
@ -32,12 +37,15 @@ The MTIA backend is implemented out of the tree, only interfaces are be defined
set_rng_state
get_rng_state
DeferredMtiaCallError
```
Streams and events
------------------
## Streams and events
```{eval-rst}
.. autosummary::
:toctree: generated
:nosignatures:
Event
Stream
```

12
docs/source/mtia.memory.rst → docs/source/mtia.memory.md
View File

@ -1,13 +1,19 @@
torch.mtia.memory
===================================
# torch.mtia.memory
The MTIA backend is implemented out of the tree, only interfaces are be defined here.
```{eval-rst}
.. automodule:: torch.mtia.memory
.. currentmodule:: torch.mtia.memory
```
```{eval-rst}
.. currentmodule:: torch.mtia.memory
```
```{eval-rst}
.. autosummary::
:toctree: generated
:nosignatures:
memory_stats
```

209
docs/source/multiprocessing.rst → docs/source/multiprocessing.md
View File

@ -1,136 +1,139 @@
:orphan:
---
orphan: true
---
.. _multiprocessing-doc:
(multiprocessing-doc)=
Multiprocessing package - torch.multiprocessing
===============================================
# Multiprocessing package - torch.multiprocessing
```{eval-rst}
.. automodule:: torch.multiprocessing
```
```{eval-rst}
.. currentmodule:: torch.multiprocessing
```
.. warning::
:::{warning}
If the main process exits abruptly (e.g. because of an incoming signal),
Python's `multiprocessing` sometimes fails to clean up its children.
It's a known caveat, so if you're seeing any resource leaks after
interrupting the interpreter, it probably means that this has just happened
to you.
:::
If the main process exits abruptly (e.g. because of an incoming signal),
Python's ``multiprocessing`` sometimes fails to clean up its children.
It's a known caveat, so if you're seeing any resource leaks after
interrupting the interpreter, it probably means that this has just happened
to you.
Strategy management
-------------------
## Strategy management
```{eval-rst}
.. autofunction:: get_all_sharing_strategies
```
```{eval-rst}
.. autofunction:: get_sharing_strategy
```
```{eval-rst}
.. autofunction:: set_sharing_strategy
```
.. _multiprocessing-cuda-sharing-details:
(multiprocessing-cuda-sharing-details)=
Sharing CUDA tensors
--------------------
## Sharing CUDA tensors
Sharing CUDA tensors between processes is supported only in Python 3, using
a ``spawn`` or ``forkserver`` start methods.
a `spawn` or `forkserver` start methods.
Unlike CPU tensors, the sending process is required to keep the original tensor
as long as the receiving process retains a copy of the tensor. The refcounting is
implemented under the hood but requires users to follow the next best practices.
.. warning::
If the consumer process dies abnormally to a fatal signal, the shared tensor
could be forever kept in memory as long as the sending process is running.
:::{warning}
If the consumer process dies abnormally to a fatal signal, the shared tensor
could be forever kept in memory as long as the sending process is running.
:::
1. Release memory ASAP in the consumer.
::
```
## Good
x = queue.get()
# do somethings with x
del x
```
## Good
x = queue.get()
# do somethings with x
del x
::
## Bad
x = queue.get()
# do somethings with x
# do everything else (producer have to keep x in memory)
```
## Bad
x = queue.get()
# do somethings with x
# do everything else (producer have to keep x in memory)
```
2. Keep producer process running until all consumers exits. This will prevent
the situation when the producer process releasing memory which is still in use
by the consumer.
::
```
## producer
# send tensors, do something
event.wait()
```
## producer
# send tensors, do something
event.wait()
::
## consumer
# receive tensors and use them
event.set()
```
## consumer
# receive tensors and use them
event.set()
```
3. Don't pass received tensors.
::
```
# not going to work
x = queue.get()
queue_2.put(x)
```
# not going to work
x = queue.get()
queue_2.put(x)
```
# you need to create a process-local copy
x = queue.get()
x_clone = x.clone()
queue_2.put(x_clone)
```
```
# putting and getting from the same queue in the same process will likely end up with segfault
queue.put(tensor)
x = queue.get()
```
::
# you need to create a process-local copy
x = queue.get()
x_clone = x.clone()
queue_2.put(x_clone)
::
# putting and getting from the same queue in the same process will likely end up with segfault
queue.put(tensor)
x = queue.get()
Sharing strategies
------------------
## Sharing strategies
This section provides a brief overview into how different sharing strategies
work. Note that it applies only to CPU tensor - CUDA tensors will always use
the CUDA API, as that's the only way they can be shared.
File descriptor - ``file_descriptor``
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
### File descriptor - `file_descriptor`
.. note::
This is the default strategy (except for macOS and OS X where it's not
supported).
:::{note}
This is the default strategy (except for macOS and OS X where it's not
supported).
:::
This strategy will use file descriptors as shared memory handles. Whenever a
storage is moved to shared memory, a file descriptor obtained from ``shm_open``
storage is moved to shared memory, a file descriptor obtained from `shm_open`
is cached with the object, and when it's going to be sent to other processes,
the file descriptor will be transferred (e.g. via UNIX sockets) to it. The
receiver will also cache the file descriptor and ``mmap`` it, to obtain a shared
receiver will also cache the file descriptor and `mmap` it, to obtain a shared
view onto the storage data.
Note that if there will be a lot of tensors shared, this strategy will keep a
large number of file descriptors open most of the time. If your system has low
limits for the number of open file descriptors, and you can't raise them, you
should use the ``file_system`` strategy.
should use the `file_system` strategy.
File system - ``file_system``
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
### File system - `file_system`
This strategy will use file names given to ``shm_open`` to identify the shared
This strategy will use file names given to `shm_open` to identify the shared
memory regions. This has a benefit of not requiring the implementation to cache
the file descriptors obtained from it, but at the same time is prone to shared
memory leaks. The file can't be deleted right after its creation, because other
@ -139,28 +142,27 @@ crash, or are killed, and don't call the storage destructors, the files will
remain in the system. This is very serious, because they keep using up the
memory until the system is restarted, or they're freed manually.
To counter the problem of shared memory file leaks, :mod:`torch.multiprocessing`
will spawn a daemon named ``torch_shm_manager`` that will isolate itself from
To counter the problem of shared memory file leaks, {mod}`torch.multiprocessing`
will spawn a daemon named `torch_shm_manager` that will isolate itself from
the current process group, and will keep track of all shared memory allocations.
Once all processes connected to it exit, it will wait a moment to ensure there
will be no new connections, and will iterate over all shared memory files
allocated by the group. If it finds that any of them still exist, they will be
deallocated. We've tested this method and it proved to be robust to various
failures. Still, if your system has high enough limits, and ``file_descriptor``
failures. Still, if your system has high enough limits, and `file_descriptor`
is a supported strategy, we do not recommend switching to this one.
Spawning subprocesses
---------------------
## Spawning subprocesses
.. note::
:::{note}
Available for Python >= 3.4.
Available for Python >= 3.4.
This depends on the ``spawn`` start method in Python's
``multiprocessing`` package.
This depends on the `spawn` start method in Python's
`multiprocessing` package.
:::
Spawning a number of subprocesses to perform some function can be done
by creating ``Process`` instances and calling ``join`` to wait for
by creating `Process` instances and calling `join` to wait for
their completion. This approach works fine when dealing with a single
subprocess but presents potential issues when dealing with multiple
processes.
@ -170,27 +172,48 @@ sequentially. If they don't, and the first process does not terminate,
the process termination will go unnoticed. Also, there are no native
facilities for error propagation.
The ``spawn`` function below addresses these concerns and takes care
The `spawn` function below addresses these concerns and takes care
of error propagation, out of order termination, and will actively
terminate processes upon detecting an error in one of them.
```{eval-rst}
.. automodule:: torch.multiprocessing.spawn
```
```{eval-rst}
.. currentmodule:: torch.multiprocessing.spawn
```
```{eval-rst}
.. autofunction:: spawn
```
```{eval-rst}
.. currentmodule:: torch.multiprocessing
```
```{eval-rst}
.. class:: SpawnContext
Returned by :func:`~spawn` when called with ``join=False``.
.. automethod:: join
```
.. This module needs to be documented. Adding here in the meantime
.. for tracking purposes
% This module needs to be documented. Adding here in the meantime
% for tracking purposes
```{eval-rst}
.. py:module:: torch.multiprocessing.pool
```
```{eval-rst}
.. py:module:: torch.multiprocessing.queue
```
```{eval-rst}
.. py:module:: torch.multiprocessing.reductions
```

308
docs/source/name_inference.rst → docs/source/name_inference.md
View File

@ -1,11 +1,12 @@
```{eval-rst}
.. currentmodule:: torch
```
.. _name_inference_reference-doc:
(name_inference_reference-doc)=
Named Tensors operator coverage
===============================
# Named Tensors operator coverage
Please read :ref:`named_tensors-doc` first for an introduction to named tensors.
Please read {ref}`named_tensors-doc` first for an introduction to named tensors.
This document is a reference for *name inference*, a process that defines how
named tensors:
@ -17,11 +18,13 @@ Below is a list of all operations that are supported with named tensors
and their associated name inference rules.
If you don't see an operation listed here, but it would help your use case, please
`search if an issue has already been filed <https://github.com/pytorch/pytorch/issues?q=is%3Aopen+is%3Aissue+label%3A%22module%3A+named+tensor%22>`_ and if not, `file one <https://github.com/pytorch/pytorch/issues/new/choose>`_.
[search if an issue has already been filed](https://github.com/pytorch/pytorch/issues?q=is%3Aopen+is%3Aissue+label%3A%22module%3A+named+tensor%22) and if not, [file one](https://github.com/pytorch/pytorch/issues/new/choose).
.. warning::
The named tensor API is experimental and subject to change.
:::{warning}
The named tensor API is experimental and subject to change.
:::
```{eval-rst}
.. csv-table:: Supported Operations
:header: API, Name inference rule
:widths: 20, 20
@ -244,226 +247,221 @@ If you don't see an operation listed here, but it would help your use case, plea
:meth:`Tensor.zero_`,None
:func:`torch.zeros`,:ref:`factory-doc`
```
.. _keeps_input_names-doc:
(keeps_input_names-doc)=
Keeps input names
^^^^^^^^^^^^^^^^^
## Keeps input names
All pointwise unary functions follow this rule as well as some other unary functions.
- Check names: None
- Propagate names: input tensor's names are propagated to the output.
::
```
>>> x = torch.randn(3, 3, names=('N', 'C'))
>>> x.abs().names
('N', 'C')
```
>>> x = torch.randn(3, 3, names=('N', 'C'))
>>> x.abs().names
('N', 'C')
(removes_dimensions-doc)=
.. _removes_dimensions-doc:
## Removes dimensions
Removes dimensions
^^^^^^^^^^^^^^^^^^
All reduction ops like :meth:`~Tensor.sum` remove dimensions by reducing
over the desired dimensions. Other operations like :meth:`~Tensor.select` and
:meth:`~Tensor.squeeze` remove dimensions.
All reduction ops like {meth}`~Tensor.sum` remove dimensions by reducing
over the desired dimensions. Other operations like {meth}`~Tensor.select` and
{meth}`~Tensor.squeeze` remove dimensions.
Wherever one can pass an integer dimension index to an operator, one can also pass
a dimension name. Functions that take lists of dimension indices can also take in a
list of dimension names.
- Check names: If :attr:`dim` or :attr:`dims` is passed in as a list of names,
check that those names exist in :attr:`self`.
- Propagate names: If the dimensions of the input tensor specified by :attr:`dim`
or :attr:`dims` are not present in the output tensor, then the corresponding names
of those dimensions do not appear in ``output.names``.
- Check names: If {attr}`dim` or {attr}`dims` is passed in as a list of names,
check that those names exist in {attr}`self`.
- Propagate names: If the dimensions of the input tensor specified by {attr}`dim`
or {attr}`dims` are not present in the output tensor, then the corresponding names
of those dimensions do not appear in `output.names`.
::
```
>>> x = torch.randn(1, 3, 3, 3, names=('N', 'C', 'H', 'W'))
>>> x.squeeze('N').names
('C', 'H', 'W')
>>> x = torch.randn(1, 3, 3, 3, names=('N', 'C', 'H', 'W'))
>>> x.squeeze('N').names
('C', 'H', 'W')
>>> x = torch.randn(3, 3, 3, 3, names=('N', 'C', 'H', 'W'))
>>> x.sum(['N', 'C']).names
('H', 'W')
>>> x = torch.randn(3, 3, 3, 3, names=('N', 'C', 'H', 'W'))
>>> x.sum(['N', 'C']).names
('H', 'W')
# Reduction ops with keepdim=True don't actually remove dimensions.
>>> x = torch.randn(3, 3, 3, 3, names=('N', 'C', 'H', 'W'))
>>> x.sum(['N', 'C'], keepdim=True).names
('N', 'C', 'H', 'W')
```
# Reduction ops with keepdim=True don't actually remove dimensions.
>>> x = torch.randn(3, 3, 3, 3, names=('N', 'C', 'H', 'W'))
>>> x.sum(['N', 'C'], keepdim=True).names
('N', 'C', 'H', 'W')
(unifies_names_from_inputs-doc)=
.. _unifies_names_from_inputs-doc:
Unifies names from inputs
^^^^^^^^^^^^^^^^^^^^^^^^^
## Unifies names from inputs
All binary arithmetic ops follow this rule. Operations that broadcast still
broadcast positionally from the right to preserve compatibility with unnamed
tensors. To perform explicit broadcasting by names, use :meth:`Tensor.align_as`.
tensors. To perform explicit broadcasting by names, use {meth}`Tensor.align_as`.
- Check names: All names must match positionally from the right. i.e., in
``tensor + other``, ``match(tensor.names[i], other.names[i])`` must be true for all
``i`` in ``(-min(tensor.dim(), other.dim()) + 1, -1]``.
`tensor + other`, `match(tensor.names[i], other.names[i])` must be true for all
`i` in `(-min(tensor.dim(), other.dim()) + 1, -1]`.
- Check names: Furthermore, all named dimensions must be aligned from the right.
During matching, if we match a named dimension ``A`` with an unnamed dimension
``None``, then ``A`` must not appear in the tensor with the unnamed dimension.
During matching, if we match a named dimension `A` with an unnamed dimension
`None`, then `A` must not appear in the tensor with the unnamed dimension.
- Propagate names: unify pairs of names from the right from both tensors to
produce output names.
For example,
::
# tensor: Tensor[ N, None]
# other: Tensor[None, C]
>>> tensor = torch.randn(3, 3, names=('N', None))
>>> other = torch.randn(3, 3, names=(None, 'C'))
>>> (tensor + other).names
('N', 'C')
```
# tensor: Tensor[ N, None]
# other: Tensor[None, C]
>>> tensor = torch.randn(3, 3, names=('N', None))
>>> other = torch.randn(3, 3, names=(None, 'C'))
>>> (tensor + other).names
('N', 'C')
```
Check names:
- ``match(tensor.names[-1], other.names[-1])`` is ``True``
- ``match(tensor.names[-2], tensor.names[-2])`` is ``True``
- Because we matched ``None`` in :attr:`tensor` with ``'C'``,
check to make sure ``'C'`` doesn't exist in :attr:`tensor` (it does not).
- Check to make sure ``'N'`` doesn't exists in :attr:`other` (it does not).
- `match(tensor.names[-1], other.names[-1])` is `True`
- `match(tensor.names[-2], tensor.names[-2])` is `True`
- Because we matched `None` in {attr}`tensor` with `'C'`,
check to make sure `'C'` doesn't exist in {attr}`tensor` (it does not).
- Check to make sure `'N'` doesn't exists in {attr}`other` (it does not).
Finally, the output names are computed with
``[unify('N', None), unify(None, 'C')] = ['N', 'C']``
`[unify('N', None), unify(None, 'C')] = ['N', 'C']`
More examples::
More examples:
# Dimensions don't match from the right:
# tensor: Tensor[N, C]
# other: Tensor[ N]
>>> tensor = torch.randn(3, 3, names=('N', 'C'))
>>> other = torch.randn(3, names=('N',))
>>> (tensor + other).names
RuntimeError: Error when attempting to broadcast dims ['N', 'C'] and dims
['N']: dim 'C' and dim 'N' are at the same position from the right but do
not match.
```
# Dimensions don't match from the right:
# tensor: Tensor[N, C]
# other: Tensor[ N]
>>> tensor = torch.randn(3, 3, names=('N', 'C'))
>>> other = torch.randn(3, names=('N',))
>>> (tensor + other).names
RuntimeError: Error when attempting to broadcast dims ['N', 'C'] and dims
['N']: dim 'C' and dim 'N' are at the same position from the right but do
not match.
# Dimensions aren't aligned when matching tensor.names[-1] and other.names[-1]:
# tensor: Tensor[N, None]
# other: Tensor[ N]
>>> tensor = torch.randn(3, 3, names=('N', None))
>>> other = torch.randn(3, names=('N',))
>>> (tensor + other).names
RuntimeError: Misaligned dims when attempting to broadcast dims ['N'] and
dims ['N', None]: dim 'N' appears in a different position from the right
across both lists.
# Dimensions aren't aligned when matching tensor.names[-1] and other.names[-1]:
# tensor: Tensor[N, None]
# other: Tensor[ N]
>>> tensor = torch.randn(3, 3, names=('N', None))
>>> other = torch.randn(3, names=('N',))
>>> (tensor + other).names
RuntimeError: Misaligned dims when attempting to broadcast dims ['N'] and
dims ['N', None]: dim 'N' appears in a different position from the right
across both lists.
```
.. note::
:::{note}
In both of the last examples, it is possible to align the tensors by names
and then perform the addition. Use {meth}`Tensor.align_as` to align
tensors by name or {meth}`Tensor.align_to` to align tensors to a custom
dimension ordering.
:::
In both of the last examples, it is possible to align the tensors by names
and then perform the addition. Use :meth:`Tensor.align_as` to align
tensors by name or :meth:`Tensor.align_to` to align tensors to a custom
dimension ordering.
(permutes_dimensions-doc)=
.. _permutes_dimensions-doc:
## Permutes dimensions
Permutes dimensions
^^^^^^^^^^^^^^^^^^^
Some operations, like :meth:`Tensor.t()`, permute the order of dimensions. Dimension names
Some operations, like {meth}`Tensor.t()`, permute the order of dimensions. Dimension names
are attached to individual dimensions so they get permuted as well.
If the operator takes in positional index :attr:`dim`, it is also able to take a dimension
name as :attr:`dim`.
If the operator takes in positional index {attr}`dim`, it is also able to take a dimension
name as {attr}`dim`.
- Check names: If :attr:`dim` is passed as a name, check that it exists in the tensor.
- Check names: If {attr}`dim` is passed as a name, check that it exists in the tensor.
- Propagate names: Permute dimension names in the same way as the dimensions that are
being permuted.
::
```
>>> x = torch.randn(3, 3, names=('N', 'C'))
>>> x.transpose('N', 'C').names
('C', 'N')
```
>>> x = torch.randn(3, 3, names=('N', 'C'))
>>> x.transpose('N', 'C').names
('C', 'N')
(contracts_away_dims-doc)=
.. _contracts_away_dims-doc:
Contracts away dims
^^^^^^^^^^^^^^^^^^^
## Contracts away dims
Matrix multiply functions follow some variant of this. Let's go through
:func:`torch.mm` first and then generalize the rule for batch matrix multiplication.
{func}`torch.mm` first and then generalize the rule for batch matrix multiplication.
For ``torch.mm(tensor, other)``:
For `torch.mm(tensor, other)`:
- Check names: None
- Propagate names: result names are ``(tensor.names[-2], other.names[-1])``.
- Propagate names: result names are `(tensor.names[-2], other.names[-1])`.
::
>>> x = torch.randn(3, 3, names=('N', 'D'))
>>> y = torch.randn(3, 3, names=('in', 'out'))
>>> x.mm(y).names
('N', 'out')
```
>>> x = torch.randn(3, 3, names=('N', 'D'))
>>> y = torch.randn(3, 3, names=('in', 'out'))
>>> x.mm(y).names
('N', 'out')
```
Inherently, a matrix multiplication performs a dot product over two dimensions,
collapsing them. When two tensors are matrix-multiplied, the contracted dimensions
disappear and do not show up in the output tensor.
:func:`torch.mv`, :func:`torch.dot` work in a similar way: name inference does not
{func}`torch.mv`, {func}`torch.dot` work in a similar way: name inference does not
check input names and removes the dimensions that are involved in the dot product:
::
```
>>> x = torch.randn(3, 3, names=('N', 'D'))
>>> y = torch.randn(3, names=('something',))
>>> x.mv(y).names
('N',)
```
>>> x = torch.randn(3, 3, names=('N', 'D'))
>>> y = torch.randn(3, names=('something',))
>>> x.mv(y).names
('N',)
Now, let's take a look at ``torch.matmul(tensor, other)``. Assume that ``tensor.dim() >= 2``
and ``other.dim() >= 2``.
Now, let's take a look at `torch.matmul(tensor, other)`. Assume that `tensor.dim() >= 2`
and `other.dim() >= 2`.
- Check names: Check that the batch dimensions of the inputs are aligned and broadcastable.
See :ref:`unifies_names_from_inputs-doc` for what it means for the inputs to be aligned.
See {ref}`unifies_names_from_inputs-doc` for what it means for the inputs to be aligned.
- Propagate names: result names are obtained by unifying the batch dimensions and removing
the contracted dimensions:
``unify(tensor.names[:-2], other.names[:-2]) + (tensor.names[-2], other.names[-1])``.
`unify(tensor.names[:-2], other.names[:-2]) + (tensor.names[-2], other.names[-1])`.
Examples::
Examples:
# Batch matrix multiply of matrices Tensor['C', 'D'] and Tensor['E', 'F'].
# 'A', 'B' are batch dimensions.
>>> x = torch.randn(3, 3, 3, 3, names=('A', 'B', 'C', 'D'))
>>> y = torch.randn(3, 3, 3, names=('B', 'E', 'F'))
>>> torch.matmul(x, y).names
('A', 'B', 'C', 'F')
```
# Batch matrix multiply of matrices Tensor['C', 'D'] and Tensor['E', 'F'].
# 'A', 'B' are batch dimensions.
>>> x = torch.randn(3, 3, 3, 3, names=('A', 'B', 'C', 'D'))
>>> y = torch.randn(3, 3, 3, names=('B', 'E', 'F'))
>>> torch.matmul(x, y).names
('A', 'B', 'C', 'F')
```
Finally, there are fused `add` versions of many matmul functions. i.e., {func}`addmm`
and {func}`addmv`. These are treated as composing name inference for i.e. {func}`mm` and
name inference for {func}`add`.
Finally, there are fused ``add`` versions of many matmul functions. i.e., :func:`addmm`
and :func:`addmv`. These are treated as composing name inference for i.e. :func:`mm` and
name inference for :func:`add`.
(factory-doc)=
.. _factory-doc:
## Factory functions
Factory functions
^^^^^^^^^^^^^^^^^
Factory functions now take a new :attr:`names` argument that associates a name
Factory functions now take a new {attr}`names` argument that associates a name
with each dimension.
::
```
>>> torch.zeros(2, 3, names=('N', 'C'))
tensor([[0., 0., 0.],
[0., 0., 0.]], names=('N', 'C'))
```
>>> torch.zeros(2, 3, names=('N', 'C'))
tensor([[0., 0., 0.],
[0., 0., 0.]], names=('N', 'C'))
(out_function_semantics-doc)=
.. _out_function_semantics-doc:
## out function and in-place variants
out function and in-place variants
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
A tensor specified as an ``out=`` tensor has the following behavior:
A tensor specified as an `out=` tensor has the following behavior:
- If it has no named dimensions, then the names computed from the operation
get propagated to it.
@ -473,13 +471,13 @@ A tensor specified as an ``out=`` tensor has the following behavior:
All in-place methods modify inputs to have names equal to the computed names
from name inference. For example:
::
```
>>> x = torch.randn(3, 3)
>>> y = torch.randn(3, 3, names=('N', 'C'))
>>> x.names
(None, None)
>>> x = torch.randn(3, 3)
>>> y = torch.randn(3, 3, names=('N', 'C'))
>>> x.names
(None, None)
>>> x += y
>>> x.names
('N', 'C')
>>> x += y
>>> x.names
('N', 'C')
```