[quant] Add dequantize.tensors #34348

jerryzh168 · 2020-03-06T02:32:59Z

Stack from ghstack:

[quant][graphmode] Make interpolate/upsample work again #35130 [quant][graphmode] Make interpolate/upsample work again
[quant][graphmode] SwapDeQuant support prim::If #35142 [quant][graphmode] SwapDeQuant support prim::If
[quant][graphmode][refactor] getGeneralOpTensorInputIndexes -> getGeneralOpTensorInputs #35141 [quant][graphmode][refactor] getGeneralOpTensorInputIndexes -> getGeneralOpTensorInputs
[quant][graphmode][refactor] swapDeQuant takes block as arugment #35135 [quant][graphmode][refactor] swapDeQuant takes block as arugment
[quant][graphmode] Fold quantized prepacking ops #35077 [quant][graphmode] Fold quantized prepacking ops
[quant] Make conv2d_prepack and linear_prepack pure #35073 [quant] Make conv2d_prepack and linear_prepack pure
[quant][graphmode] quantization support for prim::CallFunction #34855 [quant][graphmode] quantization support for prim::CallFunction
[quant][graphmode] Add quantization support for aten::cat #34346 [quant][graphmode] Add quantization support for aten::cat
[quant][graphmode] Add prim::ListConstruct to general op handling #34345 [quant][graphmode] Add prim::ListConstruct to general op handling
[quant][graphmode] quantization support for view, transpose, contiguos #34854 [quant][graphmode] quantization support for view, transpose, contiguos
[quant][graphmode] quantization support for aten::chunk #34806 [quant][graphmode] quantization support for aten::chunk
[quant][graphmode] quantization support for prim::ListUnpack #34807 [quant][graphmode] quantization support for prim::ListUnpack
[quant][graphmode] Replicate quantize node for prim::If #34804 [quant][graphmode] Replicate quantize node for prim::If
[quant][graphmode] quantization support for aten::add #34572 [quant][graphmode] quantization support for aten::add
[quant][graphmode] quantization work for prim::If #34518 [quant][graphmode] quantization work for prim::If
[quant][graphmode][fix] use observed_values_ to check values are observed #34571 [quant][graphmode][fix] use observed_values_ to check values are observed
[quant][graphmode][refactor] insertObservers for Block #34414 [quant][graphmode][refactor] insertObservers for Block
[quant][graphmode][fix] Insert dequantize before use node #34411 [quant][graphmode][fix] Insert dequantize before use node
[quant][graphmode][fix] preserve the type of original value when inserting dequant node #34349 [quant][graphmode][fix] preserve the type of original value when inserting dequant node
[quant] Add dequantize.tensors #34348 [quant] Add dequantize.tensors

Summary:
We need this function to do swap dequantize for prim::ListConstruct since
the output of prim::ListConstruct is a list of Tensors

Test Plan:
.
Reviewers:
.
Subscribers:

Tasks:

Tags:

Differential Revision: D20504454

Summary: We need this function to do swap dequantize for prim::ListConstruct since the output of prim::ListConstruct is a list of Tensors Test Plan: . Reviewers: . Subscribers: Tasks: Tags: [ghstack-poisoned]

dr-ci · 2020-03-06T02:37:46Z

💊 CircleCI build failures summary and remediations

As of commit d60b690 (more details on the Dr. CI page):

4/4 failures introduced in this PR

4 failures not recognized by patterns:

Job	Step	Status
^{caffe2_onnx_main_py3_6_clang7_ubuntu16_04_build}	^Build
^{caffe2_onnx_py2_gcc5_ubuntu16_04_build}	^Build
^{binary_linux_libtorch_2_7m_cpu_devtoolset7_shared-with-deps_build}	^{Checkout pytorch/builder repo}
^{binary_linux_manywheel_2_7mu_cpu_devtoolset7_build}	^{Checkout pytorch/builder repo}

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.

This comment has been revised 134 times.

Summary: We need this function to do swap dequantize for prim::ListConstruct since the output of prim::ListConstruct is a list of Tensors Test Plan: . Reviewers: . Subscribers: Tasks: Tags: [ghstack-poisoned]

aten/src/ATen/native/quantized/QTensor.cpp

Summary: We need this function to do swap dequantize for prim::ListConstruct since the output of prim::ListConstruct is a list of Tensors Test Plan: . Reviewers: . Subscribers: Tasks: Tags: [ghstack-poisoned]

Summary: We need this function to do swap dequantize for prim::ListConstruct since the output of prim::ListConstruct is a list of Tensors Test Plan: . Reviewers: . Subscribers: Tasks: Tags: Differential Revision: [D20504454](https://our.internmc.facebook.com/intern/diff/D20504454) [ghstack-poisoned]

smessmer · 2020-03-19T19:18:44Z

aten/src/ATen/native/quantized/QTensor.cpp

  return get_qtensorimpl(self)->quantizer()->dequantize(self);
 }

+std::vector<Tensor> dequantize_tensors_quant(TensorList tensors) {


Don't return a std::vector, return a TensorList instead (TensorList == torch::List). std::vector requires a copy behind the scenes.

I think that is not the type generated by native_functions.yaml, I looked up another function that returns Tensor[] and it says:

795:std::vector<Tensor> meshgrid(TensorList tensors); // {"schema": "aten::meshgrid(Tensor[] tensors) -> Tensor[]", "compound": "true"}

torch/_torch_docs.py

Summary: We need this function to do swap dequantize for prim::ListConstruct since the output of prim::ListConstruct is a list of Tensors Test Plan: . Reviewers: . Subscribers: Tasks: Tags: Differential Revision: [D20504454](https://our.internmc.facebook.com/intern/diff/D20504454) [ghstack-poisoned]

facebook-github-bot · 2020-03-21T08:22:48Z

This pull request has been merged in 3fa7813.

Summary: We need this function to do swap dequantize for prim::ListConstruct since the output of prim::ListConstruct is a list of Tensors Test Plan: . Reviewers: . Subscribers: Tasks: Tags: ghstack-source-id: c66bbf5 Pull Request resolved: pytorch/pytorch#34348

[quant] Add dequantize.tensors

d1147d7

Summary: We need this function to do swap dequantize for prim::ListConstruct since the output of prim::ListConstruct is a list of Tensors Test Plan: . Reviewers: . Subscribers: Tasks: Tags: [ghstack-poisoned]

Update on "[quant] Add dequantize.tensors"

74d5bf4

Summary: We need this function to do swap dequantize for prim::ListConstruct since the output of prim::ListConstruct is a list of Tensors Test Plan: . Reviewers: . Subscribers: Tasks: Tags: [ghstack-poisoned]

jerryzh168 mentioned this pull request Mar 7, 2020

[quant][graphmode][fix] Insert dequantize before use node #34411

Closed

Update on "[quant] Add dequantize.tensors"

a9084ff

Summary: We need this function to do swap dequantize for prim::ListConstruct since the output of prim::ListConstruct is a list of Tensors Test Plan: . Reviewers: . Subscribers: Tasks: Tags: [ghstack-poisoned]

jerryzh168 mentioned this pull request Mar 7, 2020

[quant][graphmode][refactor] insertObservers for Block #34414

Closed

raghuramank100 reviewed Mar 9, 2020

View reviewed changes

aten/src/ATen/native/quantized/QTensor.cpp Show resolved Hide resolved

Update on "[quant] Add dequantize.tensors"

9921674

Summary: We need this function to do swap dequantize for prim::ListConstruct since the output of prim::ListConstruct is a list of Tensors Test Plan: . Reviewers: . Subscribers: Tasks: Tags: [ghstack-poisoned]

This was referenced Mar 10, 2020

[quant][graphmode][refactor] separate code for handling subblock in CallMethod #34517

Closed

[quant][graphmode] quantization work for prim::If #34518

Closed

jerryzh168 added 2 commits March 9, 2020 20:25

Update on "[quant] Add dequantize.tensors"

7587dd6

Summary: We need this function to do swap dequantize for prim::ListConstruct since the output of prim::ListConstruct is a list of Tensors Test Plan: . Reviewers: . Subscribers: Tasks: Tags: [ghstack-poisoned]

Update on "[quant] Add dequantize.tensors"

5ea41d6

Summary: We need this function to do swap dequantize for prim::ListConstruct since the output of prim::ListConstruct is a list of Tensors Test Plan: . Reviewers: . Subscribers: Tasks: Tags: [ghstack-poisoned]

This was referenced Mar 11, 2020

[quant][graphmode][fix] use observed_values_ to check values are observed #34571

Closed

[quant][graphmode] quantization support for aten::add #34572

Closed

Update on "[quant] Add dequantize.tensors"

78c3abc

Summary: We need this function to do swap dequantize for prim::ListConstruct since the output of prim::ListConstruct is a list of Tensors Test Plan: . Reviewers: . Subscribers: Tasks: Tags: [ghstack-poisoned]

jerryzh168 added 4 commits March 18, 2020 19:43

smessmer reviewed Mar 19, 2020

View reviewed changes

jerryzh168 commented Mar 19, 2020

View reviewed changes

torch/_torch_docs.py Show resolved Hide resolved

jerryzh168 added 2 commits March 19, 2020 12:43

This was referenced Mar 19, 2020

[quant][graphmode] Fold quantized prepacking ops #35072

Merged

[quant] Make conv2d_prepack and linear_prepack pure #35073

Closed

jerryzh168 mentioned this pull request Mar 20, 2020

[quant][graphmode] Fold quantized prepacking ops #35077

Closed

jerryzh168 added 4 commits March 19, 2020 17:26

jerryzh168 mentioned this pull request Mar 20, 2020

[quant][graphmode] Make interpolate/upsample work again #35130

Closed

jerryzh168 mentioned this pull request Mar 20, 2020

[quant][graphmode][refactor] swapDeQuant takes block as arugment #35135

Closed

This was referenced Mar 21, 2020

[quant][graphmode][refactor] getGeneralOpTensorInputIndexes -> getGeneralOpTensorInputs #35141

Closed

[quant][graphmode] SwapDeQuant support prim::If #35142

Closed

facebook-github-bot closed this in 3fa7813 Mar 21, 2020

facebook-github-bot added the merged label Mar 21, 2020

facebook-github-bot deleted the gh/jerryzh168/240/head branch March 24, 2020 14:16

jerryzh168 mentioned this pull request Apr 3, 2020

[quant][graphmode] Add quantize_per_tensor.tensors #35916

Closed

mruberry added the Merged label Oct 28, 2020

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[quant] Add dequantize.tensors #34348

[quant] Add dequantize.tensors #34348

Uh oh!

jerryzh168 commented Mar 6, 2020 •

edited

Loading

Uh oh!

dr-ci bot commented Mar 6, 2020 •

edited

Loading

Uh oh!

Uh oh!

smessmer Mar 19, 2020

Uh oh!

jerryzh168 Mar 19, 2020

Uh oh!

Uh oh!

facebook-github-bot commented Mar 21, 2020

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants

[quant] Add dequantize.tensors #34348

[quant] Add dequantize.tensors #34348

Uh oh!

Conversation

jerryzh168 commented Mar 6, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

dr-ci bot commented Mar 6, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

💊 CircleCI build failures summary and remediations

4 failures not recognized by patterns:

Uh oh!

Uh oh!

smessmer Mar 19, 2020

Choose a reason for hiding this comment

Uh oh!

jerryzh168 Mar 19, 2020

Choose a reason for hiding this comment

Uh oh!

Uh oh!

facebook-github-bot commented Mar 21, 2020

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants

jerryzh168 commented Mar 6, 2020 •

edited

Loading

dr-ci bot commented Mar 6, 2020 •

edited

Loading