pytorch/torch
Elias Ellison 2381548071 add stride constraints to fallbacks (#88534)
Add stride/contiguity constraints to fallbacks so that inputs will be in the right stride permutation for the fallback kernel.

Improves perf of coat_lite_mini from 1.48415536054865 -> 2.010956856330101.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/88534
Approved by: https://github.com/ngimel
2022-11-10 01:13:44 +00:00
..
_C [Profiler] Memory profiler part 1: Gradient identification (#86802) 2022-11-08 23:53:13 +00:00
_C_flatbuffer
_decomp Fix meta for aten.upsample_bilinear2d.vec (#88158) 2022-11-02 16:58:29 +00:00
_dispatch
_dynamo [Dynamo] Fix Tensor.T trace (#88642) 2022-11-09 23:44:30 +00:00
_inductor add stride constraints to fallbacks (#88534) 2022-11-10 01:13:44 +00:00
_lazy
_prims Add min cut partitioner for AOT+nvFuser (#88204) 2022-11-09 12:56:55 +00:00
_prims_common [Reland] Fix primTorch compute_elementwise_output_strides (#88525) 2022-11-05 05:42:07 +00:00
_refs Revert "[primTorch] Improve narrow and narrow_copy: refs, tests, docs (#87045)" 2022-11-09 20:48:32 +00:00
_subclasses Mark as_strided_ as supporting SymInt in C++ (#88674) 2022-11-08 18:45:05 +00:00
amp Add AutocastHPU support (#84927) 2022-10-12 19:37:16 +00:00
ao [ao] qconfig.py fix public v private (#87515) 2022-11-09 22:30:03 +00:00
autograd Make Python op registration work with torchdeploy/multipy (#87162) 2022-11-03 12:56:44 +00:00
backends Add mem efficient backend flag (#87946) 2022-10-28 15:51:10 +00:00
contrib
cpu Add correct __all__ for torch.distributed and torch.cuda submodules (#85702) 2022-10-10 19:15:24 +00:00
csrc [13/N] Update gather with CPU/CUDA implementations (#86409) 2022-11-09 22:11:40 +00:00
cuda Deprecate TypedStorage, its derived classes, and all of their public methods (#85303) 2022-11-08 18:11:01 +00:00
distributed [nn] add remove_duplicate flag to named_parameters (#759) (#88090) 2022-11-09 00:09:20 +00:00
distributions Fix typos used in documents under torch directory (#88483) 2022-11-08 01:33:36 +00:00
fft Fix typos used in documents under torch directory (#88300) 2022-11-02 09:38:13 +00:00
futures
fx [fx] Fix GraphModule.print_readable() (#88730) 2022-11-09 21:39:48 +00:00
jit prepare removal of deprecated functionality in torch.testing (#87969) 2022-11-02 14:04:48 +00:00
legacy
lib Declare public dependencies on libshm (#82694) 2022-10-07 00:01:25 +00:00
linalg Fix typos used in documents under torch directory (#88300) 2022-11-02 09:38:13 +00:00
masked [Python] refactor slices on sorted (#86995) 2022-10-25 04:07:19 +00:00
monitor
multiprocessing Deprecate TypedStorage, its derived classes, and all of their public methods (#85303) 2022-11-08 18:11:01 +00:00
nested Implement a constructor for nested_tensor that is similar to torch.tensor() (#88213) 2022-11-08 00:03:18 +00:00
nn Delete stub file to enable mypy check (#4649) (#88701) 2022-11-09 20:29:34 +00:00
onnx [ONNX] Remove the INT64_MAX magic numbers (#88341) 2022-11-03 20:18:36 +00:00
optim Publicly expose _LRScheduler to LRScheduler (#88503) 2022-11-07 21:15:10 +00:00
package Deprecate TypedStorage, its derived classes, and all of their public methods (#85303) 2022-11-08 18:11:01 +00:00
profiler [Profiler] Memory profiler part 1: Gradient identification (#86802) 2022-11-08 23:53:13 +00:00
quantization [ao] qconfig.py fix public v private (#87515) 2022-11-09 22:30:03 +00:00
signal Reimplement Kaiser window (#87330) 2022-10-27 21:01:01 +00:00
sparse
special Adding multigammaln ref and fix arange (#85153) 2022-09-20 17:52:56 +00:00
testing Revert "[3/n] Thread PG: add threaded PG implementation (#88627)" 2022-11-09 22:38:41 +00:00
utils [fix] MathBits: serialization (#88182) 2022-11-09 17:15:12 +00:00
__config__.py
__future__.py
__init__.py Deprecate TypedStorage, its derived classes, and all of their public methods (#85303) 2022-11-08 18:11:01 +00:00
_appdirs.py
_classes.py
_deploy.py Deprecate TypedStorage, its derived classes, and all of their public methods (#85303) 2022-11-08 18:11:01 +00:00
_jit_internal.py
_linalg_utils.py Remove deprecated torch.lstsq (#70980) 2022-09-23 00:16:55 +00:00
_lobpcg.py Fix typos used in documents under torch directory (#88300) 2022-11-02 09:38:13 +00:00
_lowrank.py
_meta_registrations.py OpOverload is_view (#88722) 2022-11-09 19:03:12 +00:00
_namedtensor_internals.py
_ops.py OpOverload is_view (#88722) 2022-11-09 19:03:12 +00:00
_python_dispatcher.py [PolishComment] Polish code comment, revelant->relevant (#85238) 2022-09-19 19:43:14 +00:00
_six.py
_sources.py
_storage_docs.py
_tensor_docs.py Revert "[primTorch] Improve narrow and narrow_copy: refs, tests, docs (#87045)" 2022-11-09 20:48:32 +00:00
_tensor_str.py Disable Current Modes when printing Tensor (#88344) 2022-11-04 00:45:35 +00:00
_tensor.py [fix] MathBits: serialization (#88182) 2022-11-09 17:15:12 +00:00
_torch_docs.py Revert "[primTorch] Improve narrow and narrow_copy: refs, tests, docs (#87045)" 2022-11-09 20:48:32 +00:00
_utils_internal.py
_utils.py [fix] MathBits: serialization (#88182) 2022-11-09 17:15:12 +00:00
_VF.py
_vmap_internals.py
_weights_only_unpickler.py Revert "[fix] allow saving python attr on Tensor and Parameter via torch.save (#81616)" 2022-11-07 18:51:16 +00:00
abi-check.cpp
CMakeLists.txt Delete torch::deploy from pytorch core (#85953) 2022-10-06 07:20:16 +00:00
custom_class_detail.h
custom_class.h
extension.h
functional.py Fix typos used in documents under torch directory (#88300) 2022-11-02 09:38:13 +00:00
hub.py
library.h Make Python op registration work with torchdeploy/multipy (#87162) 2022-11-03 12:56:44 +00:00
library.py Make Python op registration work with torchdeploy/multipy (#87162) 2022-11-03 12:56:44 +00:00
overrides.py Deprecate TypedStorage, its derived classes, and all of their public methods (#85303) 2022-11-08 18:11:01 +00:00
py.typed
quasirandom.py
random.py
README.txt
return_types.py Add __all__ to torch.utils submodules (#85331) 2022-09-27 14:45:26 +00:00
script.h
serialization.py Deprecate TypedStorage, its derived classes, and all of their public methods (#85303) 2022-11-08 18:11:01 +00:00
storage.py Deprecate TypedStorage, its derived classes, and all of their public methods (#85303) 2022-11-08 18:11:01 +00:00
torch_version.py
types.py improve annotations (#86105) 2022-10-05 10:33:26 +00:00

Note [TH abstraction violation]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

TH/THC provide some hpp headers, which are proper C++ headers rather than
C headers.  These headers serve double duty as *internal implementation
detail* headers, whose contents should largely not be used by external
clients.

Ideally, we would not install these headers at all; instead, you should
use public functions (in headers like `THTensor.h`, NOT `THTensor.hpp`)
to manipulate these structs.  However, there are a few places
in torch/csrc where we violate this abstraction.  They are marked with
a pointer to this note.  Each of those sites will have to be refactored
when we refactor the guts of THTensor and related structures.