[Quant][fx] Define native backend_config_dict for linear and conv #74636

andrewor14 · 2022-03-23T20:42:55Z

Stack from ghstack (oldest at bottom):

-> [Quant][fx] Define native backend_config_dict for linear and conv #74636

Summary: This commit changes how quantization patterns for linear
and conv are set up in prepare. Previously, these were set up
through ConvReluQuantizeHandler and LinearReLUQuantizeHandler.
After this commit, however, these were set up through the
corresponding entries in the native backend_config_dict,
rendering the above quantize handlers no longer necessary.
In future commits, we will do the same for the remaining ops.

Test Plan:
python test/test_quantization.py TestQuantizeFx
python test/test_quantization.py TestQuantizeFxOps

Reviewers: jerryzh168

Subscribers: jerryzh168

Differential Revision: D35225680

Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 [ghstack-poisoned]

facebook-github-bot · 2022-03-23T20:43:01Z

🔗 Helpful links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/74636
📄 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

💊 CI failures summary and remediations

As of commit 466605a (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).

Please report bugs/suggestions to the (internal) Dr. CI Users group.

Click here to manually regenerate this comment.

Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 ghstack-source-id: 02b2bde Pull Request resolved: #74636

…nd conv" Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 [ghstack-poisoned]

Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 ghstack-source-id: f0d8b0f Pull Request resolved: #74636

torch/ao/quantization/fx/backend_config/native.py

jerryzh168

looks great, thanks!

jerryzh168 · 2022-03-23T21:53:10Z

there is some error in numeric suite, which I already fixed in my PR, should I land mine first? it might have some merge conflict with your PR

andrewor14 · 2022-03-23T22:03:07Z

Yes that's fine. I was expecting conflicts when I moved the dtype_config

…nd conv" Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 [ghstack-poisoned]

Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 ghstack-source-id: 78ca5d7 Pull Request resolved: #74636

…nd conv" Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 [ghstack-poisoned]

Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 ghstack-source-id: 22569ae Pull Request resolved: #74636

…nd conv" Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 [ghstack-poisoned]

Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 ghstack-source-id: c845963 Pull Request resolved: #74636

…nd conv" Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 [ghstack-poisoned]

Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 ghstack-source-id: 87f0733 Pull Request resolved: #74636

andrewor14 · 2022-03-29T19:23:33Z

@andrewor14 has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

…nd conv" Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 Differential Revision: [D35225680](https://our.internmc.facebook.com/intern/diff/D35225680) [ghstack-poisoned]

Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 ghstack-source-id: a901bb9 Pull Request resolved: #74636

andrewor14 · 2022-03-30T04:29:36Z

@andrewor14 has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

…nd conv" Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 Differential Revision: [D35225680](https://our.internmc.facebook.com/intern/diff/D35225680) [ghstack-poisoned]

Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 ghstack-source-id: d1b7073 Pull Request resolved: #74636

…nd conv" Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 Differential Revision: [D35225680](https://our.internmc.facebook.com/intern/diff/D35225680) [ghstack-poisoned]

andrewor14 · 2022-03-31T02:46:53Z

@andrewor14 has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

…nd conv" Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 Differential Revision: [D35225680](https://our.internmc.facebook.com/intern/diff/D35225680) [ghstack-poisoned]

Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 ghstack-source-id: 95e5680 Pull Request resolved: #74636

andrewor14 · 2022-03-31T04:37:20Z

@andrewor14 has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 ghstack-source-id: 95e5680 Pull Request resolved: #74636

…nd conv" Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 Differential Revision: [D35225680](https://our.internmc.facebook.com/intern/diff/D35225680) [ghstack-poisoned]

Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 ghstack-source-id: 7acba10 Pull Request resolved: #74636

andrewor14 · 2022-03-31T14:34:58Z

@andrewor14 has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

torch/ao/quantization/fx/backend_config/native.py

Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 ghstack-source-id: 7acba10 Pull Request resolved: #74636

…nd conv" Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 Differential Revision: [D35225680](https://our.internmc.facebook.com/intern/diff/D35225680) [ghstack-poisoned]

Summary: This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168 Subscribers: jerryzh168 ghstack-source-id: 220f25b Pull Request resolved: #74636

andrewor14 · 2022-04-02T16:46:23Z

@andrewor14 has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

…4636) Summary: Pull Request resolved: #74636 This commit changes how quantization patterns for linear and conv are set up in prepare. Previously, these were set up through ConvReluQuantizeHandler and LinearReLUQuantizeHandler. After this commit, however, these were set up through the corresponding entries in the native backend_config_dict, rendering the above quantize handlers no longer necessary. In future commits, we will do the same for the remaining ops. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Imported from OSS Reviewed By: jerryzh168, ngimel Differential Revision: D35225680 fbshipit-source-id: 4a79f63a11fce46701eb17aaf3619c1e827d72a4

github-actions · 2022-04-04T14:07:54Z

Hey @andrewor14.
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.

facebook-github-bot added the cla signed label Mar 23, 2022

facebook-github-bot added the module: fx label Mar 23, 2022

andrewor14 requested a review from jerryzh168 March 23, 2022 20:43

jerryzh168 reviewed Mar 23, 2022

View reviewed changes

torch/ao/quantization/fx/backend_config/native.py Outdated Show resolved Hide resolved

jerryzh168 approved these changes Mar 23, 2022

View reviewed changes

jerryzh168 reviewed Mar 31, 2022

View reviewed changes

torch/ao/quantization/fx/backend_config/native.py Show resolved Hide resolved

pytorchmergebot closed this in ee9335a Apr 4, 2022

andrewor14 deleted the gh/andrewor14/14/head branch April 4, 2022 14:31

andrewor14 added release notes: quantization release notes category topic: improvements topic category labels Apr 4, 2022

WBobby mentioned this pull request Aug 17, 2022

Add ROCm5.2.3/AMDGPU support for PyTorch WBobby/pytorch#2

Closed

[Quant][fx] Define native backend_config_dict for linear and conv #74636

[Quant][fx] Define native backend_config_dict for linear and conv #74636

Uh oh!

Conversation

andrewor14 commented Mar 23, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

facebook-github-bot commented Mar 23, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful links

💊 CI failures summary and remediations

Uh oh!

Uh oh!

jerryzh168 left a comment

Choose a reason for hiding this comment

Uh oh!

jerryzh168 commented Mar 23, 2022

Uh oh!

andrewor14 commented Mar 23, 2022

Uh oh!

andrewor14 commented Mar 29, 2022

Uh oh!

andrewor14 commented Mar 30, 2022

Uh oh!

andrewor14 commented Mar 31, 2022

Uh oh!

andrewor14 commented Mar 31, 2022

Uh oh!

andrewor14 commented Mar 31, 2022

Uh oh!

Uh oh!

andrewor14 commented Apr 2, 2022

Uh oh!

github-actions bot commented Apr 4, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

andrewor14 commented Mar 23, 2022 •

edited

Loading

facebook-github-bot commented Mar 23, 2022 •

edited

Loading