[Lite Interpreter] [Lite Interpreter] Operator registration migrate from manual to selective build by iseeyuan · Pull Request #35426 · pytorch/pytorch

iseeyuan · 2020-03-25T22:29:33Z

Stack from ghstack:

[Lite Interpreter] [Lite Interpreter] Operator registration migrate from manual to selective build #35426 [Lite Interpreter] Add aten op registration and use op names without "_" prefix

Use selective build with the full set of operators (vs. manually register each used op with "_" prefix).

Currently "-c pt.disable_gen_tracing=1" must be used to strip the tracing part of code-gen ops, so that the mobile code can build without dependency on full jit.

Differential Revision: D20408831

NOTE FOR REVIEWERS: This PR has internal Facebook specific changes or comments, please review them on Phabricator!

…"_" prefix Use selective build with the full set of operators (vs. manually register each used op with "_" prefix). Currently "-c pt.disable_gen_tracing=1" must be used to strip the tracing part of code-gen ops, so that the mobile code can build without dependency on full jit. Differential Revision: [D20408831](https://our.internmc.facebook.com/intern/diff/D20408831/) **NOTE FOR REVIEWERS**: This PR has internal Facebook specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D20408831/)! [ghstack-poisoned]

dr-ci · 2020-03-25T22:39:49Z

💊 CircleCI build failures summary and remediations

As of commit 1545f22 (more details on the Dr. CI page):

✅ None of the build failures appear to be your fault 💚

1/1 broken upstream at merge base 14ce500 on Apr 09 from 2:45pm to 5:52pm PDT (10 commits; 477f1c0 - c5662dd)
Please rebase on the viable/strict branch (expand for instructions)

If your commit is newer than viable/strict, you can try basing on an older, stable commit:
```
git fetch https://github.com/pytorch/pytorch viable/strict
git rebase --onto FETCH_HEAD $(git merge-base origin/master HEAD)
```
If your commit is older than viable/strict:
```
git fetch https://github.com/pytorch/pytorch viable/strict
git rebase FETCH_HEAD
```
Check out the recency history of this "viable master" tracking branch.

🚧 1 upstream failure:

These were probably caused by upstream breakages:

pytorch_xla_linux_xenial_py3_6_clang7_test on Apr 09 from 2:45pm to 5:52pm PDT (10 commits; 477f1c0 - c5662dd)
- 🔁 rerun

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.

See how this bot performed.

This comment has been revised 22 times.

…es without "_" prefix" Use selective build with the full set of operators (vs. manually register each used op with "_" prefix). Currently "-c pt.disable_gen_tracing=1" must be used to strip the tracing part of code-gen ops, so that the mobile code can build without dependency on full jit. Differential Revision: [D20408831](https://our.internmc.facebook.com/intern/diff/D20408831/) **NOTE FOR REVIEWERS**: This PR has internal Facebook specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D20408831/)! [ghstack-poisoned]

…tive build Pull Request resolved: #35426 Use selective build with the full set of operators (vs. manually register each used op with "_" prefix). Currently "-c pt.disable_gen_tracing=1" must be used to strip the tracing part of code-gen ops, so that the mobile code can build without dependency on full jit. Lite interpreter relies on JIT operator dispatch. In future we still need JIT operator dispatch dispatch ops that are not registered in c10. Currently the selective build is for c10/aten dispatch in BUCK. There is JIT selective code-gen in OSS but not ported to BUCK yet. This diff is also porting the selective code-gen in BUCK. * The selected op list is passed to gen_jit_dispatch.py. * The list passed to gen_jit_dispatch is the top-level ops (USED_PT_OPS) only, because the selective c10/aten dispatch already registered other ops that are called from the top-level ops. ghstack-source-id: 101084371 Differential Revision: [D20408831](https://our.internmc.facebook.com/intern/diff/D20408831/) **NOTE FOR REVIEWERS**: This PR has internal Facebook specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D20408831/)!

…es without "_" prefix" Use selective build with the full set of operators (vs. manually register each used op with "_" prefix). Currently "-c pt.disable_gen_tracing=1" must be used to strip the tracing part of code-gen ops, so that the mobile code can build without dependency on full jit. Differential Revision: [D20408831](https://our.internmc.facebook.com/intern/diff/D20408831/) **NOTE FOR REVIEWERS**: This PR has internal Facebook specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D20408831/)! [ghstack-poisoned]

…tive build Pull Request resolved: #35426 Use selective build with the full set of operators (vs. manually register each used op with "_" prefix). Currently "-c pt.disable_gen_tracing=1" must be used to strip the tracing part of code-gen ops, so that the mobile code can build without dependency on full jit. Lite interpreter relies on JIT operator dispatch. In future we still need JIT operator dispatch dispatch ops that are not registered in c10. Currently the selective build is for c10/aten dispatch in BUCK. There is JIT selective code-gen in OSS but not ported to BUCK yet. This diff is also porting the selective code-gen in BUCK. * The selected op list is passed to gen_jit_dispatch.py. * The list passed to gen_jit_dispatch is the top-level ops (USED_PT_OPS) only, because the selective c10/aten dispatch already registered other ops that are called from the top-level ops. ghstack-source-id: 101565513 Differential Revision: [D20408831](https://our.internmc.facebook.com/intern/diff/D20408831/) **NOTE FOR REVIEWERS**: This PR has internal Facebook specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D20408831/)!

…es without "_" prefix" Use selective build with the full set of operators (vs. manually register each used op with "_" prefix). Currently "-c pt.disable_gen_tracing=1" must be used to strip the tracing part of code-gen ops, so that the mobile code can build without dependency on full jit. Differential Revision: [D20408831](https://our.internmc.facebook.com/intern/diff/D20408831/) **NOTE FOR REVIEWERS**: This PR has internal Facebook specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D20408831/)! [ghstack-poisoned]

…tive build Pull Request resolved: #35426 Use selective build with the full set of operators (vs. manually register each used op with "_" prefix). Lite interpreter relies on JIT operator dispatch. In future we still need JIT operator dispatch dispatch ops that are not registered in c10. Currently the selective build is for c10/aten dispatch in BUCK. There is JIT selective code-gen in OSS but not ported to BUCK yet. This diff is also porting the selective code-gen in BUCK. * The selected op list is passed to gen_jit_dispatch.py. * The list passed to gen_jit_dispatch is the top-level ops (USED_PT_OPS) only, because the selective c10/aten dispatch already registered other ops that are called from the top-level ops. ghstack-source-id: 101885215 Differential Revision: [D20408831](https://our.internmc.facebook.com/intern/diff/D20408831/) **NOTE FOR REVIEWERS**: This PR has internal Facebook specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D20408831/)!

facebook-github-bot · 2020-04-10T10:08:45Z

This pull request has been merged in 7fcf8b0.

ezyang · 2020-04-10T14:56:13Z

This broke clang-format

Summary: Pull Request resolved: #36393 Fix the clang-format CI error caused in #35426 Test Plan: Imported from OSS Differential Revision: D20964854 Pulled By: iseeyuan fbshipit-source-id: 97f2ba1e006cac0f33b223315263b0b84c24cb15

…tive build (pytorch#35426) Summary: Pull Request resolved: pytorch#35426 Use selective build with the full set of operators (vs. manually register each used op with "_" prefix). Lite interpreter relies on JIT operator dispatch. In future we still need JIT operator dispatch dispatch ops that are not registered in c10. Currently the selective build is for c10/aten dispatch in BUCK. There is JIT selective code-gen in OSS but not ported to BUCK yet. This diff is also porting the selective code-gen in BUCK. * The selected op list is passed to gen_jit_dispatch.py. * The list passed to gen_jit_dispatch is the top-level ops (USED_PT_OPS) only, because the selective c10/aten dispatch already registered other ops that are called from the top-level ops. ghstack-source-id: 101885215 (Note: this ignores all push blocking failures!) Test Plan: 1. In Python, run torch.jit.export_opnames(scripted_M_mod) 2. Append the operator names into fbcode/caffe2/pt_ops.bzl and the BUCK target. 3. Run ``` buck run xplat/caffe2/fb/lite_predictor:lite_predictor_bi -- --model=/home/myuan/temp/bi_pytext_0315.bc --input_dims "1,4" --input_type int64 --pytext_len=4 ``` Should provide expected results. In addition, the size of the generated code for JIT registration, for example, ```register_aten_ops_0.cpp```, should be significantly reduced (from ~250 KB to ~80KB). The non-selected op registration schema are still kept, but the registration functor is replaced by ```DUMMY_OPERATION``` Reviewed By: ljk53 Differential Revision: D20408831 fbshipit-source-id: ec75dd762c4613aeda3b2094f5dad11804dc9492

…ch#36393) Summary: Pull Request resolved: pytorch#36393 Fix the clang-format CI error caused in pytorch#35426 Test Plan: Imported from OSS Differential Revision: D20964854 Pulled By: iseeyuan fbshipit-source-id: 97f2ba1e006cac0f33b223315263b0b84c24cb15

…"_" prefix Use selective build with the full set of operators (vs. manually register each used op with "_" prefix). Currently "-c pt.disable_gen_tracing=1" must be used to strip the tracing part of code-gen ops, so that the mobile code can build without dependency on full jit. Differential Revision: [D20408831](https://our.internmc.facebook.com/intern/diff/D20408831/) **NOTE FOR REVIEWERS**: This PR has internal Facebook specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D20408831/)! ghstack-source-id: 100876252 Pull Request resolved: pytorch/pytorch#35426

…tive build Pull Request resolved: pytorch/pytorch#35426 Use selective build with the full set of operators (vs. manually register each used op with "_" prefix). Lite interpreter relies on JIT operator dispatch. In future we still need JIT operator dispatch dispatch ops that are not registered in c10. Currently the selective build is for c10/aten dispatch in BUCK. There is JIT selective code-gen in OSS but not ported to BUCK yet. This diff is also porting the selective code-gen in BUCK. * The selected op list is passed to gen_jit_dispatch.py. * The list passed to gen_jit_dispatch is the top-level ops (USED_PT_OPS) only, because the selective c10/aten dispatch already registered other ops that are called from the top-level ops. ghstack-source-id: 101851352 Differential Revision: [D20408831](https://our.internmc.facebook.com/intern/diff/D20408831/) **NOTE FOR REVIEWERS**: This PR has internal Facebook specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D20408831/)!

…tive build (pytorch#35426) Summary: Pull Request resolved: pytorch#35426 Use selective build with the full set of operators (vs. manually register each used op with "_" prefix). Lite interpreter relies on JIT operator dispatch. In future we still need JIT operator dispatch dispatch ops that are not registered in c10. Currently the selective build is for c10/aten dispatch in BUCK. There is JIT selective code-gen in OSS but not ported to BUCK yet. This diff is also porting the selective code-gen in BUCK. * The selected op list is passed to gen_jit_dispatch.py. * The list passed to gen_jit_dispatch is the top-level ops (USED_PT_OPS) only, because the selective c10/aten dispatch already registered other ops that are called from the top-level ops. ghstack-source-id: 101885215 (Note: this ignores all push blocking failures!) Test Plan: 1. In Python, run torch.jit.export_opnames(scripted_M_mod) 2. Append the operator names into fbcode/caffe2/pt_ops.bzl and the BUCK target. 3. Run ``` buck run xplat/caffe2/fb/lite_predictor:lite_predictor_bi -- --model=/home/myuan/temp/bi_pytext_0315.bc --input_dims "1,4" --input_type int64 --pytext_len=4 ``` Should provide expected results. In addition, the size of the generated code for JIT registration, for example, ```register_aten_ops_0.cpp```, should be significantly reduced (from ~250 KB to ~80KB). The non-selected op registration schema are still kept, but the registration functor is replaced by ```DUMMY_OPERATION``` Reviewed By: ljk53 Differential Revision: D20408831 fbshipit-source-id: ec75dd762c4613aeda3b2094f5dad11804dc9492

…ch#36393) Summary: Pull Request resolved: pytorch#36393 Fix the clang-format CI error caused in pytorch#35426 Test Plan: Imported from OSS Differential Revision: D20964854 Pulled By: iseeyuan fbshipit-source-id: 97f2ba1e006cac0f33b223315263b0b84c24cb15

iseeyuan requested a review from apaszke as a code owner March 25, 2020 22:29

facebook-github-bot added the oncall: jit Add this issue/PR to JIT oncall triage queue label Mar 25, 2020

iseeyuan mentioned this pull request Mar 25, 2020

Selective build for JIT code generation #35427

Closed

This was referenced Apr 6, 2020

[Lite Interpreter] Remove register_mobile_ops.cpp #36050

Closed

[Lite Interpreter] Put mobile-used prim operators in a new file #36051

Closed

iseeyuan added 5 commits April 6, 2020 20:52

iseeyuan changed the title ~~[Lite Interpreter] Add aten op registration and use op names without "_" prefix~~ [Lite Interpreter] [Lite Interpreter] Operator registration migrate from manual to selective build Apr 9, 2020

facebook-github-bot closed this in 7fcf8b0 Apr 10, 2020

facebook-github-bot added the merged label Apr 10, 2020

iseeyuan mentioned this pull request Apr 10, 2020

Fix the clang-format error caused in register prim ops change. #36393

Closed

facebook-github-bot deleted the gh/iseeyuan/61/head branch April 13, 2020 14:16

mruberry added the Merged label Oct 28, 2020

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[Lite Interpreter] [Lite Interpreter] Operator registration migrate from manual to selective build#35426

[Lite Interpreter] [Lite Interpreter] Operator registration migrate from manual to selective build#35426
iseeyuan wants to merge 8 commits intogh/iseeyuan/61/basefrom
gh/iseeyuan/61/head

iseeyuan commented Mar 25, 2020 •

edited

Loading

Uh oh!

dr-ci Bot commented Mar 25, 2020 •

edited

Loading

Uh oh!

facebook-github-bot commented Apr 10, 2020

Uh oh!

ezyang commented Apr 10, 2020

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

iseeyuan commented Mar 25, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

dr-ci Bot commented Mar 25, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

💊 CircleCI build failures summary and remediations

🚧 1 upstream failure:

Uh oh!

facebook-github-bot commented Apr 10, 2020

Uh oh!

ezyang commented Apr 10, 2020

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

iseeyuan commented Mar 25, 2020 •

edited

Loading

dr-ci Bot commented Mar 25, 2020 •

edited

Loading