Add stateful XNNPack deconvolution2d operator to torch. #43233
Conversation
|
This pull request was exported from Phabricator. Differential Revision: D23184249 |
facebook-github-bot
added
the
oncall: jit
💊 CI failures summary and remediationsAs of commit 3b17871 (more details on the Dr. CI page): 💚 💚 Looks good so far! There are no failures yet. 💚 💚 This comment was automatically generated by Dr. CI (expand for details).Follow this link to opt-out of these comments for your Pull Requests.Please report bugs/suggestions on the GitHub issue tracker or post in the (internal) Dr. CI Users group. This comment has been revised 38 times. |
|
This pull request was exported from Phabricator. Differential Revision: D23184249 |
mitchellspryn
force-pushed
the
export-D23184249
branch
from
cdbfde0 to
423c08f
Compare
mitchellspryn
force-pushed
the
export-D23184249
branch
from
423c08f to
e278bd5
Compare
|
This pull request was exported from Phabricator. Differential Revision: D23184249 |
|
This pull request was exported from Phabricator. Differential Revision: D23184249 |
mitchellspryn
force-pushed
the
export-D23184249
branch
from
e278bd5 to
8ace18e
Compare
|
This pull request was exported from Phabricator. Differential Revision: D23184249 |
mitchellspryn
force-pushed
the
export-D23184249
branch
from
8ace18e to
1e9b8ab
Compare
|
This pull request was exported from Phabricator. Differential Revision: D23184249 |
mitchellspryn
force-pushed
the
export-D23184249
branch
from
1e9b8ab to
942c583
Compare
|
This pull request was exported from Phabricator. Differential Revision: D23184249 |
mitchellspryn
force-pushed
the
export-D23184249
branch
from
942c583 to
53b90d3
Compare
|
This pull request was exported from Phabricator. Differential Revision: D23184249 |
mitchellspryn
force-pushed
the
export-D23184249
branch
from
53b90d3 to
45ee7ca
Compare
|
This pull request was exported from Phabricator. Differential Revision: D23184249 |
mitchellspryn
force-pushed
the
export-D23184249
branch
from
45ee7ca to
2ab4019
Compare
Summary: Pull Request resolved: pytorch#43233 XNNPack is already being used for the convolution2d operation. Add the ability for it to be used with transpose convolution. Test Plan: buck run caffe2/test:xnnpack_integration Differential Revision: D23184249 fbshipit-source-id: 4172afa331bbac7db1b4bc9ca2023f30fb9a1210
|
This pull request was exported from Phabricator. Differential Revision: D23184249 |
mitchellspryn
force-pushed
the
export-D23184249
branch
from
2ab4019 to
3b17871
Compare
|
This pull request has been merged in b630c18. |
| graph(%a, %w, %b, %stride:int[], %padding:int[], %dilation:int[], | ||
| %transposed:bool, %output_padding:int[], %groups:int, %benchmark:bool, | ||
| %deterministic:bool, %cudnn_enabled:bool): |
There was a problem hiding this comment.
FYI: This pattern is deprecated, and your code might fail
|
nit: General convention for the transposed convolution is |
|
@mitchellspryn #43199 fixes the pattern, please take a look in case it messes up your parts |
4 participants
Summary:
XNNPack is already being used for the convolution2d operation. Add the
ability for it to be used with transpose convolution.
Test Plan:
Manually generate networks that contain the convolution2dT layer.
Verify that the current algorithm and the XNNPack algorithm give the same
results for a variety of inputs and parameter combinations.
Differential Revision: D23184249