torchdynamo: add convolution pointwise(unary) fusion kernel #86581

XiaobingSuper · 2022-10-10T08:39:59Z

Stack from ghstack (oldest at bottom):

Support unary fusion of Convolution with:

relu
sigmoid
tanh
hardswish
leaky_relu
hardtanh
gelu

[ghstack-poisoned]

pytorch-bot · 2022-10-10T08:40:02Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/86581

📄 Preview Python docs built from this PR
📄 Preview C++ docs built from this PR
❓ Need help or want to give feedback on the CI? Visit our office hours

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures, 4 Pending

As of commit abaafbf:
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[ghstack-poisoned]

jgong5 · 2022-10-11T11:55:30Z

aten/src/ATen/native/mkldnn/Conv.cpp

+    IntArrayRef stride,
+    IntArrayRef dilation,
+    int64_t groups,
+    std::string attr,


Use c10::string_view instead?

Yes, changed.

jgong5 · 2022-10-11T11:56:40Z

aten/src/ATen/native/mkldnn/Conv.cpp

+    int64_t groups,
+    std::string attr,
+    std::vector<c10::optional<at::Scalar>> scalars,
+    c10::optional<std::string> algorithm) {


ditto: c10::optional<c10::string_view>

cc @chunyuan-w

jgong5 · 2022-10-11T11:58:21Z

aten/src/ATen/native/mkldnn/Conv.cpp

+    IntArrayRef dilation,
+    int64_t groups,
+    std::string attr,
+    std::vector<c10::optional<at::Scalar>> scalars,


torch::List

jgong5 · 2022-10-11T12:01:02Z

test/test_mkldnn_fusion.py

@@ -13,7 +13,6 @@

 FUSION_GROUP = 'prim::TensorExprGroup'

-
 class EltwiseFusionOp(NamedTuple):


PointwisePostOp is more accurate?

Yes, I change it.

Support unary fusion of Convolution with: - relu - sigmoid - tanh - hardswish - leaky_relu - hardtanh - gelu [ghstack-poisoned]

EikanWang · 2022-10-13T02:21:23Z

@jansel , could you help review this PR for TorchInductor?

EikanWang · 2022-10-13T12:08:04Z

@jansel , I noticed that you approved the other two PRs. May I know if you have any concerns about this PR?

jansel · 2022-10-13T16:03:50Z

@jansel , I noticed that you approved the other two PRs. May I know if you have any concerns about this PR?

Nope just missed it. Sorry, too many notifications!

EikanWang · 2022-10-13T23:58:10Z

@pytorchbot merge

pytorchmergebot · 2022-10-13T23:59:38Z

Merge started

Your change will be merged once all checks pass (ETA 0-4 Hours).

Learn more about merging in the wiki.

Questions? Feedback? Please reach out to the PyTorch DevX Team

Advanced Debugging

Check the merge workflow status
here

pytorchmergebot · 2022-10-13T23:59:41Z

Merge failed

Reason: Command git -C /home/runner/work/pytorch/pytorch cherry-pick -x 21de60fc3dbadb75901e4c163bbe1d9848f8cb88 returned non-zero exit code 1

Auto-merging aten/src/ATen/native/mkldnn/Conv.cpp
CONFLICT (content): Merge conflict in aten/src/ATen/native/mkldnn/Conv.cpp
Auto-merging aten/src/ATen/native/mkldnn/Linear.cpp
error: could not apply 21de60fc3d... torchdynamo: add convolution pointwise(unary) fusion kernel
hint: After resolving the conflicts, mark them with
hint: "git add/rm <pathspec>", then run
hint: "git cherry-pick --continue".
hint: You can instead skip this commit with "git cherry-pick --skip".
hint: To abort and get back to the state before "git cherry-pick",
hint: run "git cherry-pick --abort".

Details for Dev Infra team

Raised by workflow job

EikanWang · 2022-10-14T00:01:24Z

@XiaobingSuper , could you please rebase the PR and resolve the conflict?

Support unary fusion of Convolution with: - relu - sigmoid - tanh - hardswish - leaky_relu - hardtanh - gelu [ghstack-poisoned]

XiaobingSuper · 2022-10-14T01:22:11Z

@EikanWang , code rebased.

Support unary fusion of Convolution with: - relu - sigmoid - tanh - hardswish - leaky_relu - hardtanh - gelu [ghstack-poisoned]

EikanWang · 2022-10-15T01:49:17Z

@pytorchbot merge

pytorchmergebot · 2022-10-15T01:50:58Z

Merge started

Your change will be merged once all checks pass (ETA 0-4 Hours).

Learn more about merging in the wiki.

Questions? Feedback? Please reach out to the PyTorch DevX Team

Advanced Debugging

Check the merge workflow status
here

github-actions · 2022-10-15T01:51:46Z

Hey @XiaobingSuper.
You've committed this PR, but it does not have both a 'release notes: ...' and 'topics: ...' label. Please add one of each to the PR. The 'release notes: ...' label should represent the part of PyTorch that this PR changes (fx, autograd, distributed, etc) and the 'topics: ...' label should represent the kind of PR it is (not user facing, new feature, bug fix, perf improvement, etc). The list of valid labels can be found here for the 'release notes: ...' and here for the 'topics: ...'.
For changes that are 'topic: not user facing' there is no need for a release notes label.

torchdynamo: add convolution eltwise fusion kernel

8883ad4

[ghstack-poisoned]

XiaobingSuper mentioned this pull request Oct 10, 2022

torchdynamo: add convolution pointwise(binary) fusion kernel #86582

Closed

XiaobingSuper requested review from chunyuan-w, jgong5 and EikanWang October 10, 2022 08:40

pytorchbot added the open source label Oct 10, 2022

XiaobingSuper mentioned this pull request Oct 10, 2022

torchdynamo: add linear pointwise(binary) fusion kernel #86583

Closed

Update on "torchdynamo: add convolution eltwise fusion kernel"

e236c8e

[ghstack-poisoned]

XiaobingSuper changed the title ~~torchdynamo: add convolution eltwise fusion kernel~~ torchdynamo: add convolution pointwise(unary) fusion kernel Oct 11, 2022

jgong5 requested changes Oct 11, 2022

View reviewed changes

Update on "torchdynamo: add convolution pointwise(unary) fusion kernel"

11fd97d

Support unary fusion of Convolution with: - relu - sigmoid - tanh - hardswish - leaky_relu - hardtanh - gelu [ghstack-poisoned]

XiaobingSuper requested a review from jgong5 October 12, 2022 04:37

jgong5 approved these changes Oct 12, 2022

View reviewed changes

pytorch-bot bot added the ciflow/trunk Trigger trunk jobs on your pull request label Oct 12, 2022

Update on "torchdynamo: add convolution pointwise(unary) fusion kernel"

ab9582e

Support unary fusion of Convolution with: - relu - sigmoid - tanh - hardswish - leaky_relu - hardtanh - gelu [ghstack-poisoned]

jgong5 approved these changes Oct 12, 2022

View reviewed changes

EikanWang requested a review from jansel October 13, 2022 02:19

jansel approved these changes Oct 13, 2022

View reviewed changes

EikanWang added the intel This tag is for PR from Intel label Oct 13, 2022

Update on "torchdynamo: add convolution pointwise(unary) fusion kernel"

b41d6ab

Support unary fusion of Convolution with: - relu - sigmoid - tanh - hardswish - leaky_relu - hardtanh - gelu [ghstack-poisoned]

Update on "torchdynamo: add convolution pointwise(unary) fusion kernel"

abaafbf

Support unary fusion of Convolution with: - relu - sigmoid - tanh - hardswish - leaky_relu - hardtanh - gelu [ghstack-poisoned]

pytorchmergebot added the Merged label Oct 15, 2022

pytorchmergebot closed this in 9a7a49b Oct 15, 2022

facebook-github-bot deleted the gh/XiaobingSuper/13/head branch June 8, 2023 15:00

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

torchdynamo: add convolution pointwise(unary) fusion kernel #86581

torchdynamo: add convolution pointwise(unary) fusion kernel #86581

XiaobingSuper commented Oct 10, 2022 •

edited

Loading

pytorch-bot bot commented Oct 10, 2022 •

edited

Loading

jgong5 Oct 11, 2022

XiaobingSuper Oct 12, 2022

jgong5 Oct 11, 2022

jgong5 Oct 11, 2022

XiaobingSuper Oct 12, 2022

jgong5 Oct 11, 2022

XiaobingSuper Oct 12, 2022

jgong5 Oct 11, 2022

XiaobingSuper Oct 12, 2022

EikanWang commented Oct 13, 2022

EikanWang commented Oct 13, 2022

jansel commented Oct 13, 2022

EikanWang commented Oct 13, 2022

pytorchmergebot commented Oct 13, 2022

pytorchmergebot commented Oct 13, 2022

EikanWang commented Oct 14, 2022

XiaobingSuper commented Oct 14, 2022

EikanWang commented Oct 15, 2022

pytorchmergebot commented Oct 15, 2022

github-actions bot commented Oct 15, 2022

		@@ -13,7 +13,6 @@

		FUSION_GROUP = 'prim::TensorExprGroup'


		class EltwiseFusionOp(NamedTuple):

torchdynamo: add convolution pointwise(unary) fusion kernel #86581

torchdynamo: add convolution pointwise(unary) fusion kernel #86581

Conversation

XiaobingSuper commented Oct 10, 2022 • edited Loading

pytorch-bot bot commented Oct 10, 2022 • edited Loading

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/86581

✅ No Failures, 4 Pending

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

EikanWang commented Oct 13, 2022

EikanWang commented Oct 13, 2022

jansel commented Oct 13, 2022

EikanWang commented Oct 13, 2022

pytorchmergebot commented Oct 13, 2022

Merge started

pytorchmergebot commented Oct 13, 2022

Merge failed

EikanWang commented Oct 14, 2022

XiaobingSuper commented Oct 14, 2022

EikanWang commented Oct 15, 2022

pytorchmergebot commented Oct 15, 2022

Merge started

github-actions bot commented Oct 15, 2022

XiaobingSuper commented Oct 10, 2022 •

edited

Loading

pytorch-bot bot commented Oct 10, 2022 •

edited

Loading