New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[quant] Add a new fused MovingAvg Obs + FakeQuant operator (GPU) #61589

Closed

supriyar wants to merge 15 commits into gh/supriyar/238/base from gh/supriyar/238/head

Contributor

supriyar commented

•

Stack from ghstack:

Summary:
Custom GPU implementation that does the observer + calculate qparams calculation on GPU.
It calls the aten fake_quant_per_tensor/channel functions to perform the fake quant step.

Test Plan:
python test/test_quantization.py TestFusedObsFakeQuant

Reviewers:

Subscribers:

Tasks:

Tags:

Differential Revision: D29682761


          [quant] Add a new fused MovingAvg Obs + FakeQuant operator (GPU)

d7f5e1f

Summary:
Custom GPU implementation that does the observer + calculate qparams calculation on GPU.
It calls the aten fake_quant_per_tensor/channel functions to perform the fake quant step.

Test Plan:
python test/test_quantization.py TestFusedObsFakeQuant

Reviewers:

Subscribers:

Tasks:

Tags:

[ghstack-poisoned]

supriyar requested a review from ezyang as a code owner

July 13, 2021 18:09

supriyar mentioned this pull request

[quant] Add a new fused MovingAvg Obs + FakeQuant operator(CPU) #61570

Closed

facebook-github-bot added the cla signed label

Contributor

facebook-github-bot commented

•

💊 CI failures summary and remediations

As of commit 29ef20f (more details on the Dr. CI page and at hud.pytorch.org/pr/61589):

1/1 failures introduced in this PR

🕵️ 1 new failure recognized by patterns

The following CI failures do not appear to be due to upstream breakages:

pytorch_linux_xenial_cuda11_1_cudnn8_py3_gcc7_build (1/1)

Step: "Build" (full log | diagnosis details | 🔁 rerun)

Jul 20 23:48:12 rm: cannot remove '/var/lib/jenkins/sccache_error.log': No such file or directory

Jul 20 23:48:12 ++++ extract_trap_cmd
Jul 20 23:48:12 ++++ printf '%s\n' ''
Jul 20 23:48:12 +++ printf '%s\n' cleanup
Jul 20 23:48:12 ++ trap -- '
Jul 20 23:48:12 cleanup' EXIT
Jul 20 23:48:12 ++ [[ pytorch-linux-xenial-cuda11.1-cudnn8-py3-gcc7-build != *pytorch-win-* ]]
Jul 20 23:48:12 ++ which sccache
Jul 20 23:48:12 ++ sccache --stop-server
Jul 20 23:48:12 ++ true
Jul 20 23:48:12 ++ rm /var/lib/jenkins/sccache_error.log
Jul 20 23:48:12 rm: cannot remove '/var/lib/jenkins/sccache_error.log': No such file or directory
Jul 20 23:48:12 ++ true
Jul 20 23:48:12 ++ [[ -n '' ]]
Jul 20 23:48:12 ++ [[ pytorch-linux-xenial-cuda11.1-cudnn8-py3-gcc7-build == *rocm* ]]
Jul 20 23:48:12 ++ SCCACHE_ERROR_LOG=/var/lib/jenkins/sccache_error.log
Jul 20 23:48:12 ++ SCCACHE_IDLE_TIMEOUT=1200
Jul 20 23:48:12 ++ RUST_LOG=sccache::server=error
Jul 20 23:48:12 ++ sccache --start-server
Jul 20 23:48:12 sccache: Starting the server...
Jul 20 23:48:12 ++ sccache --zero-stats
Jul 20 23:48:12 Compile requests                      0

1 job timed out:

pytorch_linux_xenial_cuda11_1_cudnn8_py3_gcc7_build

Preview docs built from this PR

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 to the (internal) Dr. CI Users group.

Click here to manually regenerate this comment.

supriyar added a commit that referenced this pull request


          [quant] Add a new fused MovingAvg Obs + FakeQuant operator (GPU)

Summary:
Custom GPU implementation that does the observer + calculate qparams calculation on GPU.
It calls the aten fake_quant_per_tensor/channel functions to perform the fake quant step.

Test Plan:
python test/test_quantization.py TestFusedObsFakeQuant

Reviewers:

Subscribers:

Tasks:

Tags:

ghstack-source-id: 56af3d77b4a309b132c75a9a3cb5a5d4c73a3d45
Pull Request resolved: #61589

supriyar marked this pull request as draft

July 13, 2021 18:09

supriyar removed the request for review from ezyang

July 13, 2021 18:09

Contributor Author

supriyar commented

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


          Update on "[quant] Add a new fused MovingAvg Obs + FakeQuant operator…

0363b19

… (GPU)"

Summary:
Custom GPU implementation that does the observer + calculate qparams calculation on GPU.
It calls the aten fake_quant_per_tensor/channel functions to perform the fake quant step.

Test Plan:
python test/test_quantization.py TestFusedObsFakeQuant

Reviewers:

Subscribers:

Tasks:

Tags:

Differential Revision: [D29682761](https://our.internmc.facebook.com/intern/diff/D29682761)

[ghstack-poisoned]

supriyar added a commit that referenced this pull request


          [quant] Add a new fused MovingAvg Obs + FakeQuant operator (GPU)

6be4cba

Summary:
Custom GPU implementation that does the observer + calculate qparams calculation on GPU.
It calls the aten fake_quant_per_tensor/channel functions to perform the fake quant step.

Test Plan:
python test/test_quantization.py TestFusedObsFakeQuant

Reviewers:

Subscribers:

Tasks:

Tags:

ghstack-source-id: 9c8ae9a757add69ce4f937122c7eb940d1b337a7
Pull Request resolved: #61589


          Update on "[quant] Add a new fused MovingAvg Obs + FakeQuant operator…

d24bc3e

… (GPU)"

Summary:
Custom GPU implementation that does the observer + calculate qparams calculation on GPU.
It calls the aten fake_quant_per_tensor/channel functions to perform the fake quant step.

Test Plan:
python test/test_quantization.py TestFusedObsFakeQuant

Reviewers:

Subscribers:

Tasks:

Tags:

Differential Revision: [D29682761](https://our.internmc.facebook.com/intern/diff/D29682761)

[ghstack-poisoned]

Contributor Author

supriyar commented

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


          Update on "[quant] Add a new fused MovingAvg Obs + FakeQuant operator…

05f4181

… (GPU)"

Summary:
Custom GPU implementation that does the observer + calculate qparams calculation on GPU.
It calls the aten fake_quant_per_tensor/channel functions to perform the fake quant step.

Test Plan:
python test/test_quantization.py TestFusedObsFakeQuant

Reviewers:

Subscribers:

Tasks:

Tags:

Differential Revision: [D29682761](https://our.internmc.facebook.com/intern/diff/D29682761)

[ghstack-poisoned]

Contributor Author

supriyar commented

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


          Update on "[quant] Add a new fused MovingAvg Obs + FakeQuant operator…

32adfd8

… (GPU)"

Summary:
Custom GPU implementation that does the observer + calculate qparams calculation on GPU.
It calls the aten fake_quant_per_tensor/channel functions to perform the fake quant step.

Test Plan:
python test/test_quantization.py TestFusedObsFakeQuant

Reviewers:

Subscribers:

Tasks:

Tags:

Differential Revision: [D29682761](https://our.internmc.facebook.com/intern/diff/D29682761)

[ghstack-poisoned]

supriyar added a commit that referenced this pull request


          [quant] Add a new fused MovingAvg Obs + FakeQuant operator (GPU)

505883c

Summary:
Custom GPU implementation that does the observer + calculate qparams calculation on GPU.
It calls the aten fake_quant_per_tensor/channel functions to perform the fake quant step.

Test Plan:
python test/test_quantization.py TestFusedObsFakeQuant

Reviewers:

Subscribers:

Tasks:

Tags:

ghstack-source-id: 3efcb78e27aa1e27c6bdc74d5b59bedc95fe2ea7
Pull Request resolved: #61589


          Update on "[quant] Add a new fused MovingAvg Obs + FakeQuant operator…

49d6a50

… (GPU)"

Summary:
Custom GPU implementation that does the observer + calculate qparams calculation on GPU.
It calls the aten fake_quant_per_tensor/channel functions to perform the fake quant step.

Test Plan:
python test/test_quantization.py TestFusedObsFakeQuant

Reviewers:

Subscribers:

Tasks:

Tags:

Differential Revision: [D29682761](https://our.internmc.facebook.com/intern/diff/D29682761)

[ghstack-poisoned]

supriyar added a commit that referenced this pull request


          [quant] Add a new fused MovingAvg Obs + FakeQuant operator (GPU)

e43bc13

Summary:
Custom GPU implementation that does the observer + calculate qparams calculation on GPU.
It calls the aten fake_quant_per_tensor/channel functions to perform the fake quant step.

Test Plan:
python test/test_quantization.py TestFusedObsFakeQuant

Reviewers:

Subscribers:

Tasks:

Tags:

ghstack-source-id: 7f24c4d993cc6165db21ce848bf82f08e48b42e9
Pull Request resolved: #61589

Contributor Author

supriyar commented

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


          Update on "[quant] Add a new fused MovingAvg Obs + FakeQuant operator…

c6eaa73

… (GPU)"

Summary:
Custom GPU implementation that does the observer + calculate qparams calculation on GPU.
It calls the aten fake_quant_per_tensor/channel functions to perform the fake quant step.

Test Plan:
python test/test_quantization.py TestFusedObsFakeQuant

Reviewers:

Subscribers:

Tasks:

Tags:

Differential Revision: [D29682761](https://our.internmc.facebook.com/intern/diff/D29682761)

[ghstack-poisoned]

supriyar added a commit that referenced this pull request


          [quant] Add a new fused MovingAvg Obs + FakeQuant operator (GPU)

ae1c16d

Summary:
Custom GPU implementation that does the observer + calculate qparams calculation on GPU.
It calls the aten fake_quant_per_tensor/channel functions to perform the fake quant step.

Test Plan:
python test/test_quantization.py TestFusedObsFakeQuant

Reviewers:

Subscribers:

Tasks:

Tags:

ghstack-source-id: 014cccdbc36229ef30f0febd8620069be63e3bee
Pull Request resolved: #61589

Contributor Author

supriyar commented

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


          Update on "[quant] Add a new fused MovingAvg Obs + FakeQuant operator…

44ba893

… (GPU)"

Summary:
Custom GPU implementation that does the observer + calculate qparams calculation on GPU.
It calls the aten fake_quant_per_tensor/channel functions to perform the fake quant step.

Test Plan:
python test/test_quantization.py TestFusedObsFakeQuant

Reviewers:

Subscribers:

Tasks:

Tags:

Differential Revision: [D29682761](https://our.internmc.facebook.com/intern/diff/D29682761)

[ghstack-poisoned]

supriyar added a commit that referenced this pull request


          [quant] Add a new fused MovingAvg Obs + FakeQuant operator (GPU)

bd37f6a

Summary:
Custom GPU implementation that does the observer + calculate qparams calculation on GPU.
It calls the aten fake_quant_per_tensor/channel functions to perform the fake quant step.

Test Plan:
python test/test_quantization.py TestFusedObsFakeQuant

Reviewers:

Subscribers:

Tasks:

Tags:

ghstack-source-id: 169406b47f75de4e0a2893185a7f2ebc099605b2
Pull Request resolved: #61589

supriyar mentioned this pull request

[quant] Create FusedMovingAvgObsFakeQuantize for QAT #61691

Closed

Contributor Author

supriyar commented

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


          Update on "[quant] Add a new fused MovingAvg Obs + FakeQuant operator…

7e5a3f4

… (GPU)"

Summary:
Custom GPU implementation that does the observer + calculate qparams calculation on GPU.
It calls the aten fake_quant_per_tensor/channel functions to perform the fake quant step.

Test Plan:
python test/test_quantization.py TestFusedObsFakeQuant

Reviewers:

Subscribers:

Tasks:

Tags:

Differential Revision: [D29682761](https://our.internmc.facebook.com/intern/diff/D29682761)

[ghstack-poisoned]

Contributor Author

supriyar commented

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

supriyar requested review from vkuzo, raghuramank100, jerryzh168 and HDCharles

July 15, 2021 17:17


          Update on "[quant] Add a new fused MovingAvg Obs + FakeQuant operator…

8d48c52

… (GPU)"

Summary:
Custom GPU implementation that does the observer + calculate qparams calculation on GPU.
It calls the aten fake_quant_per_tensor/channel functions to perform the fake quant step.

Test Plan:
python test/test_quantization.py TestFusedObsFakeQuant

Reviewers:

Subscribers:

Tasks:

Tags:

Differential Revision: [D29682761](https://our.internmc.facebook.com/intern/diff/D29682761)

[ghstack-poisoned]

Contributor Author

supriyar commented

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

supriyar marked this pull request as ready for review

July 16, 2021 18:31


          Update on "[quant] Add a new fused MovingAvg Obs + FakeQuant operator…

35ca4c8

… (GPU)"

Summary:
Custom GPU implementation that does the observer + calculate qparams calculation on GPU.
It calls the aten fake_quant_per_tensor/channel functions to perform the fake quant step.

Test Plan:
python test/test_quantization.py TestFusedObsFakeQuant

Reviewers:

Subscribers:

Tasks:

Tags:

Differential Revision: [D29682761](https://our.internmc.facebook.com/intern/diff/D29682761)

[ghstack-poisoned]

Contributor Author

supriyar commented

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

vkuzo approved these changes

View reviewed changes


          Update on "[quant] Add a new fused MovingAvg Obs + FakeQuant operator…

… (GPU)"

Summary:
Custom GPU implementation that does the observer + calculate qparams calculation on GPU.
It calls the aten fake_quant_per_tensor/channel functions to perform the fake quant step.

Test Plan:
python test/test_quantization.py TestFusedObsFakeQuant

Reviewers:

Subscribers:

Tasks:

Tags:

Differential Revision: [D29682761](https://our.internmc.facebook.com/intern/diff/D29682761)

[ghstack-poisoned]

Contributor Author

supriyar commented

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

supriyar mentioned this pull request

[quant] Remove calls to .item() for fake_quant_on #61921

Closed


          Update on "[quant] Add a new fused MovingAvg Obs + FakeQuant operator…

333c9e0

… (GPU)"

Summary:
Custom GPU implementation that does the observer + calculate qparams calculation on GPU.
It calls the aten fake_quant_per_tensor/channel functions to perform the fake quant step.

Test Plan:
python test/test_quantization.py TestFusedObsFakeQuant

Reviewers:

Subscribers:

Tasks:

Tags:

Differential Revision: [D29682761](https://our.internmc.facebook.com/intern/diff/D29682761)

[ghstack-poisoned]

Contributor Author

supriyar commented

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

supriyar added 2 commits

July 20, 2021 13:56


          Update on "[quant] Add a new fused MovingAvg Obs + FakeQuant operator…

05c6237

… (GPU)"

Summary:
Custom GPU implementation that does the observer + calculate qparams calculation on GPU.
It calls the aten fake_quant_per_tensor/channel functions to perform the fake quant step.

Test Plan:
python test/test_quantization.py TestFusedObsFakeQuant

Reviewers:

Subscribers:

Tasks:

Tags:

Differential Revision: [D29682761](https://our.internmc.facebook.com/intern/diff/D29682761)

[ghstack-poisoned]


          Update on "[quant] Add a new fused MovingAvg Obs + FakeQuant operator…

29ef20f

… (GPU)"

Summary:
Custom GPU implementation that does the observer + calculate qparams calculation on GPU.
It calls the aten fake_quant_per_tensor/channel functions to perform the fake quant step.

Test Plan:
python test/test_quantization.py TestFusedObsFakeQuant

Reviewers:

Subscribers:

Tasks:

Tags:

Differential Revision: [D29682761](https://our.internmc.facebook.com/intern/diff/D29682761)

[ghstack-poisoned]

Contributor Author

supriyar commented

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

facebook-github-bot closed this in

afdca41

Contributor

facebook-github-bot commented

This pull request has been merged in afdca41.

facebook-github-bot added the Merged label

facebook-github-bot deleted the gh/supriyar/238/head branch

July 25, 2021 14:16

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

cla signed Merged